Subido por Luiz Ventura

PremioNobel 2019 paper w18950

Anuncio
NBER WORKING PAPER SERIES
THE MIRACLE OF MICROFINANCE? EVIDENCE FROM A RANDOMIZED EVALUATION
Esther Duflo
Abhijit Banerjee
Rachel Glennerster
Cynthia G. Kinnan
Working Paper 18950
http://www.nber.org/papers/w18950
NATIONAL BUREAU OF ECONOMIC RESEARCH
1050 Massachusetts Avenue
Cambridge, MA 02138
May 2013
This paper updates and supersedes the 2010 version, which reported results using one wave of endline
surveys. The authors wish to extend thanks to Spandana, especially Padmaja Reddy whose commitment
to understanding the impact of microfinance made this project possible, and to numerous seminar audiences
and colleagues for insightful suggestions. The Centre for Micro Finance at IFMR organized and oversaw
the experiment and the data collection. Aparna Dasika and Angela Ambroz provided excellent assistance
in Hyderabad. Justin Oliver at the Centre for Micro Finance and Annie Duflo at Initiatives for Poverty
Action shared valuable advice and logistical support. Funding for the first wave of the survey was
provided by The Vanguard Charitable Endowment Program and ICICI bank. Funding for the second
wage was provided by Spandana and J-PAL. This draft was not reviewed by the The Vanguard Charitable
Endowment Program, ICICI Bank, or Spandana. Adie Angrist, Leonardo Elias, Shehla Imran, Seema
Kacker, Tracy Li, Aditi Nagaraj and Cecilia Peluffo provided excellent research assistance. The views
expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau
of Economic Research. Datasets for both waves of data used in this paper are available at
http://www.centre-for-microfinance.org/publications/data/.
NBER working papers are circulated for discussion and comment purposes. They have not been peerreviewed or been subject to the review by the NBER Board of Directors that accompanies official
NBER publications.
© 2013 by Esther Duflo, Abhijit Banerjee, Rachel Glennerster, and Cynthia G. Kinnan. All rights
reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission
provided that full credit, including © notice, is given to the source.
The Miracle of Microfinance? Evidence from a Randomized Evaluation
Esther Duflo, Abhijit Banerjee, Rachel Glennerster, and Cynthia G. Kinnan
NBER Working Paper No. 18950
May 2013
JEL No. D21,G21,O16
ABSTRACT
This paper reports on the first randomized evaluation of the impact of introducing the standard microcredit
group-based lending product in a new market. In 2005, half of 104 slums in Hyderabad, India were
randomly selected for opening of a branch of a particular microfinance institution (Spandana) while
the remainder were not, although other MFIs were free to enter those slums. Fifteen to 18 months after
Spandana began lending in treated areas, households were 8.8 percentage points more likely to have
a microcredit loan. They were no more likely to start any new business, although they were more likely
to start several at once, and they invested more in their existing businesses. There was no effect on
average monthly expenditure per capita. Expenditure on durable goods increased in treated areas,
while expenditures on “temptation goods” declined. Three to four years after the initial expansion
(after many of the control slums had started getting credit from Spandana and other MFIs ), the probability
of borrowing from an MFI in treatment and comparison slums was the same, but on average households
in treatment slums had been borrowing for longer and in larger amounts. Consumption was still no
different in treatment areas, and the average business was still no more profitable, although we find
an increase in profits at the top end. We found no changes in any of the development outcomes that
are often believed to be affected by microfinance, including health, education, and women’s empowerment.
The results of this study are largely consistent with those of four other evaluations of similar programs
in different contexts.
Esther Duflo
Department of Economics
MIT, E52-252G
50 Memorial Drive
Cambridge, MA 02142
and NBER
[email protected]
Abhijit Banerjee
MIT
Department of Economics
E52-252d
50 Memorial Drive
Cambridge, MA 02142-1347
and NBER
[email protected]
Rachel Glennerster
Abdul Latif Jameel Poverty Action Lab
MIT Department of Economics
E60-275
Cambridge MA 02139
[email protected]
Cynthia G. Kinnan
Department of Economics
Northwestern University
2001 Sheridan Road
Evanston, IL 60208
and NBER
[email protected]
1
Introduction
Microfinance institutions (MFIs) have expanded rapidly over the last 10 to 15 years: according
to the Microcredit Summit Campaign (2012), the number of very poor families with a microloan
has grown more than 18-fold from 7.6 million in 1997 to 137.5 million in 2010.
Microcredit has generated considerable enthusiasm and hope for fast poverty alleviation. In
2006, Mohammad Yunus and the Grameen Bank were awarded the Nobel Prize for Peace, for
their contribution to the reduction in world poverty. In 2009, the Consultative Group to Assist
the Poor (CGAP), an international organization housed at the World Bank and dedicated to
accelerating financial inclusion, cited the following as contributions of microfinance for which
there was already evidence: eradication of poverty and hunger, universal primary education,
the promotion of gender equality and empowerment of women, reduction in child mortality, and
improvement in maternal health. CGAP was far from alone in its enthusiasm.
The possibility of a “win-win” opportunity, in which the poor could be given the means to
pull themselves out of poverty and microfinance organizations could make a profit (potentially a
big one, as the successful IPO of Compartamos in Mexico, or SKS in India, have demonstrated)
exerts a powerful attraction on policymakers, funding agencies, and academics alike. In the last
several years, however, the enthusiasm for microcredit has been matched by an equally strong
backlash . For instance, a November 2010 article in The New York Times, appearing in the wake
of a rash of reported suicides linked to MFI over-indebtedness, quotes Reddy Subrahmanyam,
an official in Andhra Pradesh, accusing MFIs of making “hyperprofits off the poor.” He argues
that “the industry [has] become no better than the widely despised village loan sharks it was
intended to replace.... The money lender lives in the community. At least you can burn down
his house. With these companies, it is loot and scoot” (Polgreen and Bajaj 2010). MFIs have
come under attack in India (in Andhra Pradesh, an ordinance making it difficult for them to
operate has pushed several to the brink of bankruptcy), in Latin America (with the “No Pago”
movement), and even in Bangladesh (with a standoff between Yunus and the government over
the leadership of the Grameen Bank). Not unlike credit cards companies or payday lenders in
the US, MFIs are now accused of pushing their clients into debt traps. The stellar repayment
rates, once heralded as the great success of microcredit, are now cited as examples of the MFIs’
1
unscrupulous methods.
What is striking about this debate is the relative paucity of evidence to inform it. Anecdotes
about highly successful entrepreneurs or deeply indebted borrowers tell us nothing about the
effect of microfinance on the average borrower, much less the effect of having access to it on the
average household. Even representative data about microfinance clients and non-clients cannot
identify the causal effect of microfinance access, because clients are self-selected and therefore
not comparable to non-clients. Microfinance organizations also purposely choose some villages
and not others. . Difference-in-difference estimates can control for fixed differences between
clients and non-clients, but it is likely that people who choose to join MFIs would be on different
trajectories even absent microfinance. This invalidates comparisons over time between clients
and non-clients (see Alexander-Tedeschi and Karlan 2007).
These issues make the evaluation of microcredit particularly difficult, and there is so far
no consensus among academics on its impact. For example, Pitt and Khandker (1998) use the
eligibility threshold for getting a loan from Grameen bank as a source of identifying variation
in a structural model of the impact of microcredit, and find large positive effects, especially for
women. However, Jonathan Morduch (1998), and Roodman and Morduch (2010) criticize the
approach, pointing out among other issues that there is in fact no discontinuity in the probability
to borrow at that threshold.1
As early as 1999, Morduch wrote that “the ‘win-win’ rhetoric promising poverty alleviation
with profits has moved far ahead of the evidence, and even the most fundamental claims remain
unsubstantiated.” In 2005, Beatriz Armendáriz and Morduch reiterated the same uncertainty in
their book The Economics of Microfinance, noting that the relatively few carefully conducted
longitudinal or cross-sectional impact studies yielded conclusions much more measured than
MFIs’ anecdotes would suggest, reflecting the difficulty of distinguishing the causal effect of
microcredit from selection effects. These cautions were repeated in the book’s second edition in
2010 .
Given the complexity of this identification problem, the ideal experiment to estimate the effect
1
Kaboski and Townsend (2005) use a natural experiment (the introduction of a village fund whose size is fixed
by village) to estimate the impact of the amount borrowed and find impacts on consumption, but not investment.
This is a government-provided form of credit and differs in a number of ways from the standard microcredit
product.
2
of having access to microcredit is to randomly assign microcredit to some areas, and not others,
and compare outcomes in both. Randomization ensures that, on average, the only difference
between residents is the greater ease of access to microcredit of those in the treatment area.2
In this paper we report on the first randomized evaluation of the effect of the canonical grouplending microcredit model, which targets women who may not necessarily be entrepreneurs.3
This study also follows the households over the longest period of any study (it followed households
for about three to 3.5 years after the introduction of the program in their slums areas), which is
necessary since many impacts may be only expected to surface over the medium run. A number
of recent papers have reported on subsequent randomized evaluations of similar programs in
Morocco (Crépon et al., 2011), Bosnia-Herzegovina (Augsburg et al., 2012), Mexico (Angelucci
et al., 2012) and Mongolia (Attanasio et al. 2011). We will compare their results to ours in the
last section of this paper.4
The experiment was conducted as follows. In 2005, 52 of 104 poor neighborhoods in Hyderabad were randomly selected for opening of an MFI branch by one of the fastest-growing
MFIs in the area, Spandana, while the remainder were not. Hyderabad is the fifth largest city in
India, and the capital of Andhra Pradesh, the Indian state were microcredit has expanded the
fastest. Fifteen to 18 months after the introduction of microfinance in each area, a comprehensive household survey was conducted in an average of 65 households in each neighborhood, for a
total of about 6,850 households. In the meantime, other MFIs had also started their operations
in both treatment and comparison households, but the probability of receiving an MFI loans
was still 8.8 percentage points (48%) higher in treatment areas than in comparison areas (27.1%
borrowers in treated areas versus 18.3% borrowers in comparison areas). Two years after this
first endline survey, we surveyed the same households once more. By that time, both Spandana
and other organizations had started lending in the treatment and control groups, so the fraction
of households borrowing from microcredit organizations was not significantly different (38.5% in
2
An alternative to measure the impact of borrowing is to randomize microcredit offer among applicants. This
approach was pioneered by Karlan and Zinman (2009), which uses individual randomization of the “marginal”
clients in a credit scoring model to evaluate the impact of consumer lending in South Africa, and find that access
to microcredit increases the probability of employment. Karlan and Zinman (2011) use the same approach to
measure impact of microcredit among small businesses in Manila.
3
The two studies mentioned in Footnote 2 evaluate slightly different programs: consumer lending in the case
of Karlan and Zinman (2009), and “second generation” individual liability loans to existing entrepreneurs in the
case of Karlan and Zinman (2010).
4
See Banerjee (2013) for a comprehensive summary of the recent literature on microcredit.
3
treatment and 33% in control). But households in treatment groups had larger loans and had
been borrowing for a longer time period. This second survey thus gives us an opportunity to
examine some of the longer-term impacts of microcredit access on households and businesses.
To frame the analysis, we propose a model where a household may wish to acquire lumpy
investment (a durable good, or an asset for a business ). One key result of the model is that
households who have access to microcredit may sacrifice short- or even medium-term consumption
when microcredit becomes available in order to get the durable good, or to invest in a business.
Other households may decide to expand their labor supply. Non-durable consumption may
thus initially fall, and even total consumption may not increase. Of course, if the household
has invested in a profitable business, we could eventually expect consumption to increase: this
underscores the importance of following households over a long enough period of time.
We examine the effect consumption, new business creation, business income, etc., as well
as measures of other human development outcomes such as education, health and women’s
empowerment. At the first endline, we see no difference in monthly per capita consumption and
monthly non-durable consumption. We do see significant positive impacts on the purchase of
durables. There is evidence that this is financed partly by an increase in labor supply and partly
by cutting unnecessary consumption: households have reduced expenditures on what that they
themselves describe as “temptation goods.”
Thus, in our context, microfinance plays a role in helping households make different intertemporal choices in consumption. This is not the only impact that is traditionally expected from
microfinance, however. The primary engine of growth that it is supposed to fuel is business
creation. Fifteen to 18 months after gaining access, households are no more likely to be entrepreneurs (that is, have at least one business), but they are more likely to start more than one
business, and they invest more in the businesses they do have (or the ones they start). There is
an increase in the average profits of the businesses that were already in existence before microcredit, but this is entirely due to very large increases in the upper tail. At every quantile between
the 5th and the 95th percentile, there is no difference in the profits of the businesses. The median marginal new business is both less profitable and less likely to have even one employee in
treatment than in control areas.
After three years, when microcredit is available both in treatment and control groups but
4
treatment group households have had the opportunity to borrow for a longer time, businesses
in the treatment groups have significantly more assets, and business profits are now larger for
businesses above the 85th percentile. However, the average business is still small and not very
profitable. In other words, contrary to most people’s belief, to the extent microcredit helps
businesses, it may help the larger businesses more. There is still no difference in average consumption.
We do not find any effect on any of the women’s empowerment or human development
outcomes either after 18 or 36 months. Furthermore, almost 70% of eligible households do not
have an MFI loan, preferring instead to borrow from other sources, if they borrow (and most
do).
Our results find a strong echo in the four other studies that look at similar programs in
different contexts. This gives us confidence in the robustness and external validity of our findings.
In short, microcredit is not for every household, or even most households, and it does not lead
to the miraculous social transformation some proponents have claimed. Its principal impact
seems, perhaps unsurprisingly, to allow some households to sacrifice some instantaneous utility
(temptation goods or leisure) to finance lumpy purchases, either for their home or in order to
establish or expand a business.
2
Experimental Design and Background
2.1
The Product
Until the major crisis in Indian microfinance in 2010, Spandana was one of the largest and fastest
growing microfinance organizations in India, with 1.2 million active borrowers in March 2008, up
from 520 borrowers in 1998-9, its first year of operation (MIX Market, 2009). From its birthplace
in Guntur, a dynamic city in Andhra Pradesh, it has expanded across the state and into several
others.
The basic Spandana product is the canonical group loan product, first introduced by the
Grameen Bank. A group is comprised of six to ten women, and 25-45 groups form a “center.”
Women are jointly responsible for the loans of their group. The first loan is Rs. 10,000, about $200
at market exchange rates, or $1,000 at 2007 purchasing power parity (PPP)-adjusted exchange
5
rates (World Bank, 2007).5 It takes 50 weeks to reimburse principal and interest rate; the interest
rate is 12% (non-declining balance; equivalent to a 24% APR). If all members of a group repay
their loans, they are eligible for second loans of Rs. 10,000-12,000; loan amounts increase up to
Rs. 20,000.
Unlike other microfinance organizations, Spandana does not require its clients to start a
business (or pretend to) in order to borrow: the organization recognizes that money is fungible,
and clients are left entirely free to choose the best use of the money, as long as they repay
their loan. Also unlike other microlenders, most notably Grameen, Spandana does not insist on
“transformation” in the household. Spandana is primarily a lending organization, not directly
involved in business training, financial literacy promotion, etc.
Eligibility is determined using the following criteria: clients must (a) be female,6 (b) be aged
18 to 59, (c) have resided in the same area for at least one year, (d) have valid identification
and residential proof (ration card, voter card, or electricity bill), and (e) at least 80% of women
in a group must own their home. Groups are formed by women themselves, not by Spandana.
Spandana does not determine loan eligibility by the expected productivity of the investment,
although selection into groups may screen out women who cannot convince fellow group-members
that they are likely to repay.
2.2
Experimental Design
Spandana initially selected 120 areas (identifiable neighborhoods, or bastis) in Hyderabad as
places in which they were interested in opening branches. These areas were selected based on
having no preexisting microfinance presence, and having residents who were desirable potential
borrowers: poor, but not “the poorest of the poor.” Areas with high concentrations of construction workers were avoided because they move frequently which makes them undesirable as
microfinance clients. While the selected areas are commonly referred to as “slums,” these are
permanent settlements with concrete houses and some public amenities (electricity, water, etc.).
5
In 2007 the PPP exchange rate was $1=Rs. 9.2, while the market exchange rate was $1'Rs. 50. All following
references to dollar amounts are in PPP terms unless noted otherwise.
6
Spandana also offers an individual-liability loan. Men are also eligible for individual-liability loans, and
individual borrowers must document a monthly source of income, but the other criteria are the same as for jointliability loans. 96.5% of Spandana borrowers were female in 2008 (Mix Market, 2009). Spandana introduced the
individual-liability loan in 2007; very few borrowers in our sample have individual-liability loans.
6
Within eligible neighborhoods, the largest ones were not selected for the study, since Spandana
was keen to start operations there. The population in the neighborhoods selected for the study
ranges from 46 to 555 households.
In each area, we conducted a small baseline neighborhood survey in 2005, collecting information on household composition, education, employment, asset ownership, expenditure, borrowing,
saving, and any businesses currently operated by the household or stopped within the last year.
We surveyed a total of 2,800 households in order to obtain a rapid assessment of the baseline
conditions of the neighborhoods. However, since there was no existing census, and the baseline
survey had to be conducted very rapidly to gather some information necessary for stratification before Spandana began their operations, the households were not selected randomly from a
household list: instead field officers were asked to map the area and select every nth house, with
n chosen to select 20 household per area. But this procedure was not very rigorous, and we are
not confident that the baseline is representative. Thus, the baseline survey was used as a basis
for stratification, a descriptive analysis below, and area-level characteristics are used as control
variables.7 Beyond this, we do not use the baseline survey in the analysis that follows.
After the baseline survey, but prior to randomization, sixteen areas were dropped from the
study because they were found to contain large numbers of migrant-worker households. Spandana
(like other MFIs) has a rule that loans should only be made to households who have lived in the
same community for at least one year because the organization believes that dynamic incentives
(the promise of more credit in the future) are more important in motivating repayment for these
households. The remaining 104 areas were grouped into pairs of similar neighborhoods, based
on average per capita consumption and per-household debt, and one of each pair was randomly
assigned to the treatment group.8
Table 1 uses the baseline sample to show that treatment and comparison areas did not
differ in their baseline levels of demographic, financial, or entrepreneurship characteristics in the
baseline survey. This is not surprising, since the sample was stratified according to per capita
consumption, fraction of households with debt, and fraction of households who had a business.9
7
Omitting these controls does not affect the results.
Pairs were formed to minimize the sum across pairs A, B (area A avg loan balance – area B avg loan balance)2
+ (area A per capita consumption – area B per capita consumption)2 . Within each pair one neighborhood was
randomly allocated into treatment.
9
Since the sample of households was not random at baseline, we also verify that the households surveyed at
8
7
The baseline data also provides a snapshot of households’ characteristics prior to Spandana’s
expansion, which we discuss further below.
Spandana then progressively began operating in the 52 treatment areas, between 2006 and
2007. Note that in the intervening periods, other MFIs also started their operations, both in
treatment and comparison areas, and we did nothing to stop that. We will show below that there
is still a significant difference between MFI borrowing in treatment and comparison groups.
To create a proper sampling frame for the endline, we undertook a comprehensive census of
each area in early 2007, and included a question on borrowing. The census revealed low rates
of MFI borrowing even in treatment areas, so the endline sample consisted of households whose
characteristics suggested high likelihood of having borrowed: households who had resided in
the area for at least three years and contained at least one woman aged 18 to 55. Spandana
borrowers identified in the census were oversampled, and the results presented below correct for
this oversampling so that the results are representative of the population as a whole. Since they
were not representative, baseline households were not purposely resurveyed in the follow-up.
We began the endline survey in August 2007 and ended it in April 2008. In each area, this
first endline survey was conducted at least 12 months after Spandana began disbursing loans,
and generally 15 to 18 months after. The overall sample size for the endline survey was 6,864
households.
Two years later, in 2009-2010, we undertook a second endline survey, following up on the same
households, asking the same set of questions as in 2007-2008 to insure comparability. Appendix
Table 2, Panel A shows, the re-contact rate at endline 2 for household initially interviewed at
endline 1 was very high, at 89.9% in the treatment group and 90.2% in the control group. Panel
B shows average characteristics of the recontacted versus attrited households. The samples do
not differ significantly along most dimensions. However, those who attrited had higher per capita
expenditure at endline 1, by Rs. 131 (column 1). Attritors were five percentage points less likely
to have an MFI loan at endline 1 (column 5), and 1.5 percentage points less likely to have a
business created in the one year prior to endline 1 (column 7). This is consistent with businesses
and microloans being associated with lower mobility, and higher consumption/permanent income
endline are similar in treatment and control groups, in terms of a number of characteristics which are fixed over
time (Table A1).
8
being associated with higher mobility. Panel C shows that one important characteristic differentially predicts attrition in treatment versus control, namely MFI borrowing: the attrited sample
is nine percentage points less likely than the non-attrited sample to have had an MFI loan in
treatment areas. This suggests that Spandana was effective in either targeting households that
were going to stay put, or convincing them not to leave the area.10
2.3
The context
Table 1 shows a snapshot of households from the 104 sampled areas in 2005. Recall that these
numbers need to be viewed with some caution, as the households sampled at baseline were not
necessarily representative of the area as a whole, and were not purposely resurveyed at endline.
At baseline, the average household (averaging over treatment and control areas) was a family of
five, with monthly expenditure of just under Rs. 5350, or $540 at PPP-adjusted exchange rates
($108 per capita) (World Bank, 2005). A majority of households (67%) lived in a house they
owned, and 27% in a house they rented.11 Almost all of the 7 to 11 year olds (98%), and 86% of
the 12 to 15 year olds, were in school.
There was almost no MFI borrowing in the sample areas at baseline. However, 68% of the
households had at least one outstanding loan. The average amount outstanding was Rs. 21,658
(median Rs. 11,000), and the average interest rate was 3.89% per month. Most loans were taken
from moneylenders (50%), friends or neighbors (25%), and family members (13%). Commercial
bank loans were very rare (3%).
Although business investment was not commonly named as a motive for borrowing, 24% of
households ran at least one small business at the baseline, compared to an OECD-country average
of 12% who say that they are self-employed. However, these businesses were very small. Only
7.5% had any employees; typical assets included sewing machines, tables and chairs, balances and
10
While attrition rates are comparable in treatment and comparison areas, the differential attrition according
to propensity to borrow from an MFI is potentially concerning, not only for the analysis of endline 2 data, but
possibly for endine 1 as well: endline 1 data may suffer from attrition, although we do not observe it since we do
not have a baseline. To address this concern, we have re-estimated all the regressions below with a correction for
sample selection inspired by Dinardo, Fortin and Lemieux (2010), where we re-weight the data using the inverse
of the propensity to be observed at endline 2, so that the distribution of observable characteristics (at endline 1)
among households observed at endline 2 resembles that in the entire endline 1 sample. We then apply the same
weights to endline 1 data (implicitly assuming a similar selection process between the onset of microfinance and
endline 1). The results, available upon request, are very similar to what we present here.
11
The remaining 6% had missing information to the home ownership question.
9
pushcarts, and 15% of businesses had no assets whatsoever. Average revenues were approximately
Rs. 9,900 ($980 in PPP terms) per month on average. Business income (i.e., profits) were
approximately Rs. 3,300 ($325 at PPP). Total household income, from entrepreneurship, wage
labor, irregular labor, etc. averaged approximately Rs. 4,840. Forty-two percent of working
individuals worked for a wage.
Baseline data revealed more limited use of consumption smoothing strategies other than
borrowing: 34% of the households had a savings account, and only 23% had a life insurance
policy. Almost none (0.03%) had any health insurance. Forty percent of households reported
spending Rs. 550 ($54) or more on a health shock in the last year; 50% of households who had
a sick member had to borrow for a health-related purpose.
Growth between 2005 and 2010
Table 2, shows some of the same key statistics for the endline 1 and endline 2 (EL1 and EL2)
samples in the control group.
Comparing the control baseline sample (2005) with the control households in the EL1 (2008)
and EL2 (2010) samples reveal rapid secular growth in Hyderabad over 2005-2010.12 Average
household consumption rose from Rs. 5,485 to Rs. 7,662 in 2007 and Rs. 11,497 (all expressed
in 2007 rupees). in EL2. There was a 12 percentage point increase in the likelihood the family’s
house was waterproof between baseline and EL2 (68% versus 56%). Eighty-one percent of families
owned a color TV at EL2, up 20 percentage points from two years before and 50 percentage points
from the baseline. The fraction owning a cellphone increased from 17% at baseline to 64% at
EL1 and 86% at EL2.
The percentage of households who ran at least one small business increased from 24% at
baseline to 34% at EL1 and 42% at EL2. Forty-three percent of these businesses were primarily
operated by a woman. However, the businesses remain very small: only 9% (10%) had any
employees at EL1 (EL2). Yet despite remaining very small in terms of employment, average
revenues rose from approximately Rs. 9,900 ($980 in PPP terms) per month on average at
baseline to just over Rs. 11,000 at EL1 and almost 16,000 at EL2. At EL2, business owners
12
While the comparison may not be perfect since the baseline survey was not conducted on the same sample as
the endline, the growth between EL1 and EL2 is for the same set of households, using the same survey instruments,
and thus gives us a good sense of the dynamism of this economy.
10
reported business income (profits) of almost Rs. 5,000 (~$540 at PPP), up from about Rs. 2,500
($275) at EL1. (These profit estimates do not account for the cost of the proprietors’ time.)
The fraction of households with at least one outstanding loan rose from 68% at baseline to
89% in EL1 and 90% in EL2. The use of consumption-smoothing strategies other than borrowing
also increased. From 34%, the fraction of households with a savings account skyrocketed to 82%
at EL1 and 85% at EL2, and the fraction with health insurance rose from almost 0 at baseline
to 12% at EL1 and 76% at EL2, likely due to the expansion of the government’s RSBY health
insurance program from those below the poverty line. Nonetheless, at EL1 (EL2), 64% (78%) of
households reported spending Rs. 500 or more on a health shock in the last year. The fraction
of households who had a sick member that had to borrow held fairly constant: 50% at baseline
to 53% at EL1 and 45% at EL2.
2.4
Treatment impact on MFI borrowing and borrowing from other sources
Treatment communities were randomly selected to receive Spandana branches, but other MFIs
also started operating both in treatment and comparison areas. We are interested in testing
the impact of microcredit, not only borrowing from Spandana. Table 3 Panel A shows that,
by the first endline, MFI borrowing was indeed higher in treatment than in control slums, although borrowing from other MFIs made up for part of the difference in Spandana borrowing.
Households in treatment areas are 13.3 percentage points more likely to report being Spandana
borrowers–18.5% versus 5.2% (Table 3 Panel A, column 2). The difference in the percentage
of households saying that they borrow from any MFI is 8.8 points (Table 3 Panel A, column
1), so some households who ended up borrowing from Spandana in treatment areas would have
borrowed from another MFI in the absence of the intervention. While the absolute level of total
MFI borrowing is not very high, it is about 50% higher in treatment than in comparison areas.
Columns 5 and 7 show that treatment households also report significantly more borrowing from
MFIs (and from Spandana in particular) than comparison households. Averaged over borrowers
and non-borrowers, treatment households report Rs. 1,391 more borrowing from Spandana than
do control households, and Rs. 1,355 more from all MFIs.
While both the absolute take up rate and the implicit “first stage” are relatively small,
this appears to be similar to what was found in other evaluations of the impact of access to
11
microfinance, despite the different contexts. In rural Morocco, Crépon et al. (2011) find that
the probability of having any loan from the MFI Al Amana in areas which got access to it
is 10 percentage points, whereas it is essentially zero in control, and moreover, since there is
really no other MFI, this represents the total increase in microfinance borrowing. In Mexico,
Angelucci, Karlan and Zinman (2012) find an increase in 10 percentage points in the probability
of borrowing from the MFI Compartamos in areas that got access to the lender, relative to a
base of five percentage points in the control (they don’t report the probability to borrow from
any other MFI). In Mongolia, Attanasio et al. (2011) find a much larger increase, 48 percentage
points, but this is among a sample that had already expressed interest in obtaining a loan from
the lender and formed a potential borrowing group before randomization.13
The fairly low take up rate in these difference contexts is in itself is a perhaps surprising
result, given the high levels of informal borrowing in these communities and the purported
benefits of microcredit over these alternative forms of borrowing. . In all cases, except when the
randomization was among those who had already expressed explicit interest in microcredit, only
a minority of “likely borrowers” end up borrowing.
Table 3 also displays the impact of microfinance access on other forms of borrowing. A
sizable fraction of the clients report repaying a more expensive debt as a reason to borrow from
Spandana, and we do indeed see some action on this margin, but column 3 shows that the share
of households who have some informal borrowing–defined as borrowing from family, friends,
moneylenders and goods purchased on credit–goes down by 5.2 percentage points in treatment
areas, but bank borrowing is unaffected. The point estimate of the amount borrowed from
informal sources is also negative, suggesting substitution of expensive borrowing with cheaper
MFI borrowing (an explicit objective of Spandana), and the point estimate, though insignificant,
is quite similar in absolute value to the increase in MFI borrowing (column 8). However, given the
high level of informal borrowing, this corresponds to a decline of only 2.6%: When we examine
the distribution of endline 1 informal borrowing, in Figure 1, informal borrowing is significantly
lower in treatment areas from the 30th to 65th percentiles.
After the end of the first endline, following our initial agreement with Spandana, the control
13
The last study with which we consider, Augsburg et al. (2012), is not strictly comparable to ours because
the sampling frame is made up of people who had applied for a loan. But even there the difference in borrowing
rates between treatment and control group is fairly low, only 20 percentage points.
12
slums were “released,” and Spandana was free to expand in these areas. Other MFIs also continued their expansion. However, two years later a significant difference still remained between
Spandana slums and others: Table 3 Panel B shows that 18% of the households in the treatment
slums borrowed from Spandana, against 11% in the control slums. Other MFIs continued to
expand both in the former treatment and control slums, and MFI lending overall was almost the
same in the treatment and the control group. By the second endline survey, 33.1% of households
had borrowed from an MFI in the former control slums, and 33.7% in the treatment slums. Since
lending started later in the control group, however, households in the treatment group had on
average been borrowing for longer than those in the control group, which is reflected in the fact
that they had completed more loan cycles. On average, there was a difference of 0.13 loan cycles
between the treatment and the control households at endline 2 (column 10), which is almost
unchanged from endline 1. . The key difference between treatment and control group at endline
2 is thus the length of access to microfinance. Since microfinance loans grow with each cycle,
treatment households also had larger loans. Among those who borrow, there was by the endline
2 a significant difference of Rs. 2,344 (or 14%) in the size of the loans (column 6). Since about
one third of households borrow, this translates into an (insignificant) difference of about Rs. 869
in average borrowing (column 5).
3
Theory
Since the stated goal of many MFIs is to help their client escape poverty by investing in their own
businesses, evaluations of microfinance programs (including this one) typically focus on business
investments and overall consumption per capita as key measures of success. However, to the
extent that microfinance successfully relaxes credit constraints, we may see households sacrifice
short-run non-durable consumption to invest in durable goods (either for home consumption or
for their businesses). The short-run impact (as people take the loan and then repay it) may
therefore be to reduce non-durable consumption or even overall consumption. The increase
in welfare would either come from the utility arising from the durable consumption or, in the
longer run, if the investment makes the borrower’s businesses more profitable and that feeds into
increases in consumption. This suggests that if consumption is a main outcome of interest, we
13
need to pay attention to its composition. Also, a relatively long horizon may be necessary to
determine the full effects. The simple model below clarifies this intuition in order to provide a
conceptual frame to our analysis.
3.1
Basic Model
A consumer lives for T 2 periods. We assume just for expositional convenience that T is even.
She consumes two goods which we will call non-durable and durable. The non-durable is fully
divisible and is consumed in the period it is bought. Denote non-durable consumption by cn .
The durable lasts for two periods, and yields durable services in both periods. The durable is
indivisible and costs an amount cd , and yields durable services of acd in each period. Moreover
there are no additional benefits from owning a second durable. Assume that durable services and
non-durables are perfect substitutes in the sense that the consumer’s per-period utility function
is u(c), where c = cn if she has not purchased the durable in the current or previous period and
c = cn + acd otherwise. Assume that 0 < a < 1. Therefore in the current period purchasing
the durable leads to a net loss in flow utility, but it might still be optimal because a could be
greater than 1/2. The consumer does not discount and the future and therefore maximizes total
of present and future utility.
The consumer earns a labor income of y in units of the non-durable every period and there is
no savings or investment, so the total amount y is spent every period. However, the household
has the option of borrowing up to an amount bmax for one period at a gross interest rate r. We
assume, in keeping with the microfinance application, that the person cannot borrow again till
after the loan is fully repaid. In other words, if the borrower borrows in period s, she will have
to repay in period s + 1 and can only borrow again in period s + 2. Finally we assume that the
durable costs more than the maximum possible amount of debt: cd > bmax
Given this, the consumer’s problem in each period depends just on whether she already owns
the durable and her existing stock of debt. If she owns the durable she has no reason to buy it
in the current period; if she has debt then she has to repay it in the current period and cannot
borrow until the next period.
14
3.2
Analysis of the model
The structure of this model yields a very useful simplification. In the Theoretical Appendix we
show that the consumer’s decision can be analyzed by simply looking at the decision in the first
two periods, assuming that there are no further periods. The decision in the first period will
be repeated in all subsequent odd periods and what happens in period 2 will be repeated in all
subsequent even periods.
This is very convenient because we can study the decision diagrammatically. In Figure 2, the
horizontal axis represents consumption in period 1 and the vertical axis is consumption in period
2. U U and U 0 U 0 are two potential indifference curves. They both have slope 1/δ when they
intersect the 45 degree line, OO0 at points E and E 0 . The point E represents the endowment,
the vector (y, y). The line EF , which has the slope r, represents the set of options open to the
consumer if he borrows in period 1 but does not purchase the durable. The distance along the
horizontal direction from E to F represents bmax , the maximum possible loan size. As drawn, we
are assuming that r < 1/δ, which gives the consumer a reason to borrow–the highest indifference
curve reachable on EF is typically higher that the one through E.
The other option is to buy the durable. The point A represents the case of just buying the
durable and not borrowing, i.e. it is the point (y − (1 − a)cd , y + acd ). The line segment AB
represents the set of choices for someone who borrows and buys the durable. The horizontal
distance from A to B is bmax and the slope of the line is r. As drawn, it is clear that the point
B lies on the highest indifference curve that is available and the consumer will choose both to
borrow and to buy the durable. However, her first-period consumption is still lower than at
point E. Non-durable consumption and even total consumption goes down in the first period as
a result of purchasing the durable.
However, this is not the only possibility. The point B 0 represents what happens when bmax
is higher (F 0 is the corresponding point where the consumer borrows without purchasing the
durable). In this case, borrowing and buying the durable is still the best option, but total
consumption goes up in both periods. Finally, the point B 00 represents the case where bmax is
small. F 00 is the corresponding value in the case where there is no durable purchase. In this case,
borrowing without buying the durable is the best option, and first-period consumption goes up.
Figure 3 captures the case where rδ > 1. In this case there is no reason to just borrow–the
15
line EF lies everywhere under the indifference curve through E. However, borrowing to buy the
durable still makes sense and improves welfare.
In general, more credit (weakly) increases the incentive to buy the durable relative to either
not buying but borrowing or not buying and not borrowing. To see this denote the utility of
buying the durable as vd (bmax ), and that of not buying the durable by vn (bmax ).
d
d
vd (bmax ) = max{ [u(y−(1−a)cd +b)+δu(y+cd −rb)], 0} = max{u0 (y−(1−a)cd +b)−δru0 (y+acd −rb), 0}
dbmax
db
which, by the concavity of u is always at least as large as
dvn (bmax )
dbmax
d
= max{ db
[u(y + b) + δu(y −
rb)], 0} = max{u0 (y+b)−δru(y−rb), 0}.Therefore this is also true at the point where vd (bmax ) =
vn (bmax ), which tells us that if is tells us that if at any level of bmax vd (bmax ) > vn (bmax ), then
this is also true at all higher values of bmax . In this sense, increased access to credit favors buying
the durable.
Moreover, it is evident that when the consumer switches to buying the durable as a result
of increased credit access, his borrowing must go up. Hence, compared to someone who has less
credit access, his second-period non-durable consumption, y − rb must be lower.
Result 1: Compare two people, one of whom has higher access to credit. She is more likely to
buy the durable, but her first-period total non-durable consumption and even total consumption
may be higher or lower. Her second-period non-durable consumption will be lower.
3.3
Extensions
We have so far ignored possibility of making a productive investment. Note however that the
model where there are no durables but the consumer has a choice of investing a fixed amount
(1 − a)cd in period 1 to get a return of acd in period 2 is formally identical to the model
with durables and the same reasoning applies. However, the change in interpretation makes
worth emphasizing that since a higher a means a more productive project, for a high enough a
the investment will be made even when access to credit is very limited or absent. Conversely,
increased access to credit will encourage consumers with relatively low values of a to invest.
Result 2: Increased access to credit increases the likelihood that the consumer makes a
fixed investment but reduce the average product of the projects that get implemented. Total
16
first-period consumption can go up or down with greater credit access. However, in this case the
person will have higher second-period non-durable consumption, since that is the reason for the
investment.
Next, consider a variant of the model where the consumer also has a labor supply decision.
Assume that the consumer can earn w units of non-durable consumption per unit of labor and
supplies l1 and l2 units of labor in periods 1 and 2. The disutility of labor is given by the function
v(l) which is assumed to be increasing, convex, differentiable everywhere and satisfying the Inada
condition at l = 0. The consumer now maximizes
u(y − (1 − a)cd + b + wl1 ) − v(l1 ) + δ[u(y + cd − rb + wl2 ) − v(l2 )
if she buys the durable and
u(y + b + wl1 ) − v(l1 ) + δ[u(y − rb + wl2 ) − v(l2 )]
if not.
By our assumptions about v,an interior optimum for l always exists and is given by
u0 (c) = v 0 (l).
It is evident that l is decreasing in c. Furthermore, if ul (x) = maxl {u(x + wl) − v(l)}, it is easy
to show that ul (x) inherits the concavity of u(c) and therefore Result 1 extends to this case.
In other words, improved loan access may lead to a reduction in non-durable and even total
consumption in the first period. If total consumption goes down, labor supply will go up in that
period.
Result 3: Increased access to credit can lead to an increase in labor supply in the first
period.
Finally, the assumption that durables and non-durables are perfect substitutes is convenient
for diagrammatic analysis but not essential for our results. Suppose, on the contrary, durable
consumption of cd leads to an utility equal to the service flow from the durable acd , which is
separable from the utility from non-durables. Then it is easily shown by following the same
17
argument that Result 1 will still hold. The only change is that now labor supply only depends
on non-durable consumption, and since non-durable consumption can be lower in both periods,
labor supply may be permanently raised by improved credit access.14
Result 4: If durables and non-durables are not perfect substitutes, increased access to credit
may raise labor supply in both periods.
3.4
Discussion
The main point made in the theoretical section is that increased access to credit can lead to
lowered non-durable consumption, both when the loan is taken and while it is being repaid,
and increased labor supply, potentially once again both when the loan is taken and while it is
being repaid. Durable consumption must, of course, go up if the point of the borrowing is to
buy a durable (though it may not be picked up depending on when in the borrowing cycle the
comparison is made), but not necessarily if the point is to start a business.
To interpret the results below, we consider that the period between the baseline and endline
1 corresponds to two model periods (one borrowing cycle), and the period between endline 1 and
endline 2 corresponds to the next two model periods (one borrowing cycle). This is realistic,
as the baseline happened roughly 15 to 18 months after Spandana started its operation in each
slum, and the average borrowing household had been borrowing for a quarter.
The model tells us that the second borrowing cycle can be just like the first, if there are
multiple durables to buy. In this case, we may see very little difference between those who got
credit access on the first round with those who got it later, except to the extent that the loan size
goes up from round to round–bigger loans may allow buying bigger durables. Of course, if the
durables actually last for more than two periods, those who have access to microfinance earlier
will have a larger stock of durables. On the other hand, if credit is used in both periods to invest
in a business and those businesses are in fact profitable, consumption should be higher in endline
2 for households in the initial treatment group since they are already enjoying the business
returns, while control households have yet to do so.15 Observing the dynamic of treatment effect
14
The same result also holds when instead of durables and non-durables, the consumer chooses between a
divisible consumption good and a non-divisible one (say, a wedding).
15
This is, of course, unless they borrow more and invest everything in the business, in which case we may see
higher profits, but potentially consumption that is no higher.
18
across two borrowing cycles, with one group gaining access one round later, is therefore useful
in assessing the overall impact of microcredit on poverty.
4
Results
To estimate the impact of microfinance becoming available in an area, we focus on intent to
treat (ITT) estimates; that is, simple comparisons of averages in treatment and comparison
areas, averaged over borrowers and non-borrowers. We present ITT estimates of the effect of
microfinance on businesses operated by the household; and for those who own businesses, we
examine business profits, revenue, business inputs, and the number of workers employed by the
business. (The construction of these variables is described in Appendix 2.) Each column reports
the results of a regression of the form
0
yia = α + β × T reatia + Xa γ + εi
where yia is an outcome for household i in area a, T reatia is an indicator for living in a treated
0
area, and β is the intent to treat effect. Xa is a vector of control variables, calculated as arealevel baseline values: area population, total businesses, average per capita expenditure, fraction
of household heads who are literate, and fraction of all adults who are literate.16 Standard
errors are adjusted for clustering at the area level and all regressions are weighted to correct for
oversampling of Spandana borrowers.
4.1
Consumption
Table 4 gives intent to treat estimates of the effect of microfinance on household spending.
Columns 1 and 2 of Panel A shows that there is no significant difference in total household
expenditures–either total or non-durable–per adult equivalent, between treatment and comparison households. The point estimate is essentially zero in both cases and we can reject the null
hypothesis that there was a Rs. 85 per month increase in consumption per adult equivalent
and Rs. 56 (about 6% of the average in control for consumption, and 4% for non-durable con16
Table A1 shows that treatment and comparison areas are balanced in terms of these characteristics so, as
expected, the results are very similar, although slightly less precise, if these controls are omitted.
19
sumption) increase. Thus enhanced microcredit access does not appear to be associated with a
significant increase in consumption after 15 to 18 months. Of course, this may partly be due
to the fact that relatively few people borrow, and that some in the control group borrow from
another MFI; still, even if the entire increase (decrease) in total (non-durable) consumption was
due to borrowers, these point estimates thus suggest very modest effects of borrowing.17
While there are no significant impacts on average consumption and non-durable consumption,
there are shifts in the composition of expenditure: column 4 shows that households in treatment
areas spent a statistically significant Rs. 1154 more on durables over the past year than did
households in comparison areas. Note that this is probably an underestimate of the total effect
of loans on durable purchases, since our measure would miss anyone who borrowed more than a
year before the survey (the survey was 15 to 18 months after the centers opened) and immediately
bought a durable with the loan proceeds. The most commonly purchased durables include gold
and silver, motorcycles, sarees (purchased in bulk, presumably mainly for weddings), color TVs,
fridges, rickshaws, computers and cellphones.
Consistent with the model, column 2 shows that while there was no detectable change in
non-durable spending otherwise, the increase in durable spending by treatment households was
essentially offset by reduced spending on “temptation goods” and festivals. Temptation goods
are goods that households in our baseline survey said that they would like spend less on (this
is thus the same list of goods for all households). They include alcohol, tobacco, betel leaves,
gambling, and food consumed outside the home. Spending on temptation goods is reduced by
about Rs. 9 per family per month (column 3). We also see in column 5 a large fall in festival
spending per capita in the previous year (Rs. 763, significant at the 10% level). Together, the
average drop in consumption in temptation goods and festivals over the year is Rs. 1255 per
family and per year, which is reasonably close to the average increase in durables spending of
Rs. 1154 plus the average interest difference of Rs. 325. The remaining difference of about Rs.
225 per family per year is probably matched by extra labor earnings (labor supply increases by
3.23 hours per week).18 The decrease in festival expenditures does not come from large changes
17
For total consumption, the implied IV estimate is a Rs. 113 (10/.088) or 5% increase, and for non-durable it
is a Rs. 75 (4%) decrease.
18
Rs. 1255 comes from 763+8.73*4.68*12 where 8.73 is the reduction on temptation good spending per capita
per month and 4.68 is average household size. Rs. 325 comes from 0.24*1355 where 24% is the interest paid on
the net extra MFI borrowing of Rs. 1355. Households in the treatment group also spend on average an extra Rs.
20
in large, very expensive ceremonies such as weddings (we see very few of them in the data) but
rather appears to come from declines at all levels of the distribution of spending on festivals.
This is consistent with the model, assuming that every month (a subperiod of our model), some
households get a chance to make a lumpy expenditure if they can and want to pay for it. We
then see them undertake a large expense (this is more likely in microcredit borrower households
because they have the money). The expense is paid from small, monthly cuts in the non-necessary
non-durable goods (temptation goods and festivals) and spread across both households who have
already made the lumpy purchase and households that anticipate doing so in the future.19
Panel B of Table 4 reports on the impact effects at the time of the second endline, when
both treatment and control households have access to the microfinance program. The effects on
both total per capita spending and total per capita non-durable spending (columns 1 and 2) are
negative with t-statistics around 1. Spending on temptation goods is still lower by about Rs. 10
per month (column 3), similar to endline 1, though the effect is now insignificant. The effect on
festivals is now positive but nowhere near being significant. Overall, the substantial gap between
treatment and control households in terms of avoidable non-durable spending we saw in the first
endline has shrunk to about Rs. 375 per year in endline 2. Correspondingly, there is no difference
on average in durable goods spending in endline 2 (column 4). Given that the main difference
between treatment and control households at endline 2 is that treatment households have been
borrowing for longer, this suggests that, in the second cycle, households in the treatment seem
to just repeat the first cycle with another durable (of roughly the same size), while households
in the control group also acquire a durable.
The absolute magnitude of these changes is relatively small: for instance, the Rs. 1154 of
increased durables spending at endline 1 is approximately $125 at 2007 PPP exchange rates.
However, this represents an increase of about 17% relative to total spending on durable goods in
comparison areas. Furthermore, this figure averages over non-borrowers and borrowers. If all of
this additional spending were coming from the extra 8.8% who do borrow (that is, if there were
no spillover effects to non-borrowers), then the implied increase per additional borrower would
389 per year acquiring business assets compared to households in control group but they also make Rs. 357 more
in profits (not significant) which would potentially balance it out.
19
Households do not have to spend the loan as soon as they get it. We see examples of households taking a
loan and putting it in a savings account for some time before they spend it.
21
be more than twice the level of durable goods spending in comparison areas. However, since it
is entirely possible that there are spillover or general equilibrium effects (as analyzed by Buera
et al. 2011), and effects that operate through the expectation of being able to borrow when
needed (such as reductions in precautionary savings, as documented in Thailand by Kaboski
and Townsend (2011) and in India by Fulford (2011)), or through general-equilibrium effects on
prices or wages (Giné and Townsend 2004), we will focus here on reduced-form/intent to treat
estimates.
4.2
New businesses and business outcomes
The basic version of our model describes the situation of a household deciding whether to acquire
a durable good. However, as we discuss in the extension (Section 3.3), it can also apply to the
decision to invest in a new or existing business. Most microcredit organizations seek to help
poor women become entrepreneurs or improve the profitability of their business, and it is thus
important to investigate the effect on business creation and expansion.
Panel A in Table 5 presents the results from the first endline on business outcomes. Column
1 indicates that the probability that a household starts a business is in fact not significantly
different in treatment and control areas. In comparison areas, 4.7% of households opened at
least one business in the year prior to the survey, compared to 5.6% in treated areas. However,
treatment households were somewhat more likely to have opened more than one business in the
past year, and column 2 shows that more businesses were created in treatment areas overall:
5.3 per 100 households in control versus 6.9 per 100 households in treatment. The numbers
are small because there are few businesses opened in any given year, but the treatment effect
represents a significant proportional increase (30% more business were created in treatment than
in comparison areas).
Consistent with the fact that Spandana loans only to women, and with the stated goals of
microfinance institutions, the marginal businesses tend to be female-operated: column 3 shows
that when we look at creation of businesses that are owned by women,20 we find that almost
all of the differential business creation in treatment areas is in female-operated businesses–there
20
A business is classified as owned by a woman if the first person named in response to the question “Who is
the owner of this business?” is female. Only 72 out of 2674 businesses have more than one owner. Classifying a
business as owned by a woman if any person named as the owned is female does not change the result.
22
are 0.015 percentage points more female-owned businesses in treatment than in control areas, an
increase of 58%. Households in treated areas were no more likely to report closing a business,
an event reported by 3.9% of households in treatment areas and 3.7% of the households in
comparison areas (column 4).21
Consistent with the fact that treatment households start more businesses, they invest more in
durables for the business. Since only a third of households have a business, and most businesses
use no asset whatsoever, the point estimate is small in absolute value (Rs. 389 over the last year,
or a bit less than a third of the increase in average MFI borrowing in treatment households) but
the increment in treatment is more than the total value of business durables purchased in the
last year by comparison households (Rs. 280), and is statistically significant.
The rest of the columns in the Panel A of Table 5 report on current business status and
last month’s revenues, inputs costs, and profits. In these regressions, we assign a zero to those
households who do not have a business, so these results give us the overall impact of credit
on business activities, including both the extensive and intensive margins. Consistent with the
prior results, treatment households are no more likely to have a business (summing over those
created in the last year and those created before) but they have more business assets (although
the t-statistic on the asset stock is only 1.58). The treatment effects on revenues and inputs are
both positive but insignificant.
Finally, there is an insignificant increase in business profits. Since this data includes zeros for
households who do not have a business, this answers the question of whether microcredit, as it
is often believed, increases poor households’ income by expanding their business opportunities.
The point estimate, at Rs. 357 per month corresponds to a roughly 50% increase relative to the
profits received by the average comparison household. This is thus large in proportion of profits,
but it represents only a very small increase in disposable income for an average households–recall
that the average total consumption of these households is about Rs. 7,000 per month and an
increase of Rs. 357 per month in business revenues is certainly not going to change the life of
the average person who gets access to microcredit.
That does not rule out that the businesses of some specific groups could have benefit from
21
It is possible that households not represented in our sample, such as households who had not lived in the area
for three years, may have been differentially likely to close businesses in treated areas. However, the relatively
small amount of new business creation makes general-equilibrium effects on existing businesses rather unlikely.
23
the loan. To look at this in more detail, we focus on businesses that were already in existence
before microcredit started. We do this in Table 6.22 For businesses that existed before Spandana
expanded, we find an average increase in profits of Rs. 2194 in treatment areas, which is significant and more than double the control mean. This increase, however, is entirely concentrated
in the upper tail (quantiles 95 and above), as shown in Figure 4. At every other quantile, there
is very little difference between the profits of existing businesses in treatment and control areas.
The 95th percentile of monthly profit of existing businesses is Rs. 14600 (or $1590 at PPP),
which makes them quite large and profitable businesses in this setting. The vast majority of the
small businesses make very little profits to start with, and microcredit does nothing to help them.
This absence of an effect on the average business is consistent with the results of Karlan and
Zinman (2011), who evaluate individual loans given to micro-entrepreneurs in the Philippines,
and do not find that the loans result in an increase in profits. The finding that microcredit is
most effective in helping larger businesses is contrary both to much of the rhetoric of microcredit
and the view of microcredit skeptics.
Finally, we have seen that the treatment led to some more business creation, particularly
female-owned businesses. In Figure 5, and Tables 7 and A3, we show more data on the characteristics of these new businesses. The quantile regressions in Figure 5 (profits for businesses that
did not exist at baseline) show that all businesses between the 35th and 65th percentiles have
significantly lower profits in treatment areas. Table 7, column 1 shows that the mean profit is
not significantly different across treatment and control due to the noisy data, but the median
new business in treatment areas has Rs. 1250 lower profits, significant at the 5% level (Table
7, column 2). The average new business is also significantly less likely to have employees in the
treatment areas: the proportion of the new businesses that have any employee falls from 9.4%
to only 4.5% (column 5).
These results could in principle be a combination of a treatment effect and a selection effect,
but since the effect on existing businesses suggests a treatment effect which is close to zero
for most businesses (and the point estimate is positive), the effect for new businesses is likely
due to selection–the marginal business that gets started in treatment areas is less profitable
22
In Table 5, we show that households are no more or less likely to close a business in the last year, thus there
is no sample selection induced by microfinance.
24
than the marginal business in the control areas. The hypothesis that the marginal business
which gets started is different in the treatment group gains some additional support in Appendix
Table 3, which shows a comparison of the industries of old businesses and new businesses, across
treatment and comparison areas.23 Industry is a proxy for the average scale and capital intensity
of a business, which is likely to be measured with less error than actual scale or asset use. The
industry composition of new businesses do differ. In particular, the fraction of food businesses
(tea/coffee stands, food vendors, kirana stores, and agriculture) is 8.5 percentage points (about
45%) higher among new businesses in treatment areas than among new businesses in comparison
areas, and the fraction of rickshaw/driving businesses among new businesses in treatment areas
is 5.4 (more than 50%) percentage points lower. Both these differences are significant at the
10% level. Food businesses are the least capital-intensive businesses in these areas, with assets
worth an average of just Rs. 930 (mainly dosa tawas, pots and pans, etc.). Rickshaw/driving
businesses, which require renting or owning a vehicle, are the most capital-intensive businesses,
with assets worth an average of Rs. 12,697 (the bulk of which is the cost of the vehicle). The
result that the marginal business created is less profitable in treatment than in control areas is
consistent with our model (see Result 2), but the fact that they are smaller bears some discussion.
Indeed, these households clearly do not need a loan to be able to start a business that requires
Rs. 930 worth of assets. We therefore interpret this as a labor supply effect along the lines of
Result 3: households use most of the loan to pay for something like a durable and then increase
their labor supply to pay back the loan, perhaps using a small part of the loan to buy some
inputs that they need to work.
Another explanation for both results could be that the marginal businesses are more likely
to be female owned, and are thus started in sectors where women are active, and businesses
operated by women tend in general to be less profitable, perhaps because of social constraints
on what they can do and how much effort they can devote to it.24
Panel B of Table 5 shows the results for the business performance variables at the time of the
second endline. As remarked already, by this time treatment and control households are equally
likely to have a microcredit loan, but the loan in treatment areas is bigger and borrowers have
23
Respondents could classify their businesses into 22 different types, which we grouped into the following: food,
clothing/sewing, rickshaw/driving, repair/construction, crafts vendor, and “other.”
24
This is true in our data, and also found for example in Sri Lanka by de Mel et al. (2009).
25
been borrowing for a longer time. The results follow a clear pattern, consistent with the idea
that control households now borrow at the same rate. We find no difference in business creation
in treatment and control areas. The new businesses are in the same industries in treatment and
control areas, and the negative effects at the median have disappeared (result omitted). For
the contemporaneous flow investment outcomes such as new business creation, business assets
acquired in the previous year, etc. (columns 1 through 5 ) the point estimate is very close to
zero (however the standard errors are large). On the other hand, businesses in treatment areas
have significantly larger asset stock (column 6), which reflects the cumulative effect of the past
years during which they had a chance to borrow and expand. Despite this, their profits are still
not significantly larger, though the point estimate is around 60% of the sample mean (with a
t-statistics of around 1.2). As shown in Figure 6, the positive increase is once again concentrated
in the top and bottom tails, although it starts being positive a little earlier, at the 85th percentile.
Overall, these results lead us to revise downward the role of microcredit as primarily being an
engine of escape from poverty through small business growth. Microfinance is indeed associated
with (some) business creation: in the first year, it does lead to an increase in the number of
new businesses created, particularly by women (though not in the number of households that
start a business). However, these marginal businesses are even smaller and less profitable than
the average business in the area (the vast majority of which are already small and unprofitable).
It does also lead to a greater investment in the existing businesses, and an improvement in the
profits for the most profitable of those businesses. For everyone else, business profits do not
increase, and on average microfinance does not help the businesses to grow in any significant
way. (Even after three years, there is no increase in the number of employees of businesses that
existed before Spandana started its operation.)
4.3
Labor supply
Our last theoretical result is on labor supply: access to credit can lead to an increase of labor
supply, as households who have acquired a durable good give up leisure to finance the purchase.
Table 8 shows the impact of the program on labor supply. In endline 1, adults (head and
spouse) in treatment households increase their overall labor supply by an average of 3.22 hours.
The increase occurs entirely in the households’ own businesses, and there is no increase number
26
of hours worked for wages: those hours may be much less elastic, if the households do not
fully choose them. However, unlike Augsburg et al. (2012), we do not find the increase in
teenagers’ labor supply that is sometimes feared to be a potential downside of microfinance (as
the adolescents are drawn into the business by their parents). By endline 2, as control households
have started borrowing, the difference between treatment and control disappears.
4.4
Microfinance as social revolution: education, health, and women’s empowerment?
The evidence so far suggests a different picture from the standard description of the role of
microfinance in the life of the poor: the pent-up demand for it is not overwhelming; many
households use their loan to acquire an household durable, reducing avoidable consumption to
finance it; some invest in their businesses, but this does not lead to significant growth in the
profitability of their business(es). Another staple of the microfinance literature is that because
the loans are given to women and give them a chance to start their own business, this would
lead to a more general empowerment of women in the households, and this empowerment would
in turn translate in better outcomes for everyone, including education, health, etc. (e.g. CGAP,
2009). To examine these questions, Table 9 examines the effects of access to microfinance on
measure of women’s decision-making and children’s health. Column 1 shows that women in
treatment areas were no more likely to be the primary decision-makers regarding decisions about
household spending, investment, savings, or education. Column 2 shows that even focusing
on decisions other than what food to purchase, which might be more sensitive to changes in
empowerment, does not change the finding. Column 3 shows that, among households with loans
(88% of all households), women in treatment-area households were no more likely to report being
the person in the household who decided to take the loan.
A finding of many studies of household decision-making is that an increase in women’s bargaining power leads to an increase in food and health expenditure (See Thomas (1990) and Duflo
(2003)). Thus, health investments and outcomes are interesting in their own right, and increased
spending in these areas might also demonstrate greater decision-making or bargaining power for
women. However, we find no effect on health outcomes.25 In Table 4, column 8, we find that
25
We do not find changes in food expenditure or food share either.
27
households in treatment areas spend no more on health and sanitation items than do comparison households. Table 9, column 4 shows that, among households with children, households in
treatment areas were no less likely to report that a child had a major illness in the past year.
In Table 4, we show that in endline 1, there was no impact on education expenditures either.
We also find that there is no change in the probability that children or teenagers are enrolled in
school (Table 9, columns 5 and 6), private school fees (Table 4, column 10), or in private school
versus public school enrollment (results not reported to save space).
Because there are many possible proxies for women’s empowerment, and many “social” outcomes we could consider, examining one at a time will create a multiple inference problem–out
of 20 outcomes, we expect that one would differ between treatment and control at the 5% level
of significance even if the microcredit intervention had no impact. To address this, we use the
approach of Kling et. al. (2007) to test the null hypothesis of no effect of microcredit on “social
outcomes” against the alternative that microcredit improves social outcomes. We construct an
equally weighted average of z-scores for the 16 social outcomes; this method gives us maximal
power to detect an effect on social outcomes, if such an effect is present.26 Column 7 shows
that there is no effect on the index of social outcomes (point estimate .007 standard deviations)
and we can rule out an increase of more than one twentieth of a standard deviation with 95%
confidence.
This suggests that there is no prima facie evidence that microcredit leads to important
changes in household decision-making or in social outcomes. Furthermore, this is not only
because we observe this only in the short run. Nothing major changes by endline 2: the effect
of microfinance access on the index of women empowerment is still very small and insignificant,
and anything but a small effect can still be ruled out. Recall that we are comparing households
who, by EL2, are equally likely to borrow: the main difference by EL2 is that households in
the treatment group have had greater access to microfinance for the first 18 months; this may
limit power to detect differences in the social outcomes at the community level. Nevertheless,
there are two interesting differences: First, while there is still little difference in overall education
26
The 16 outcomes we use are: indicators for women making decisions on each of food, clothing, health, home
purchase and repair, education, durable goods, gold and silver, investment; levels of spending on school tuition,
fees, and other education expenses; medical expenditure; teenage girls’ and teenage boys’ school enrollment; and
counts of female children under one year and one to two years old. We selected these outcomes because they
would likely be affected by changes in women’s bargaining power within the household.
28
expenditure, there is by endline 2 a significant difference in how much is spent on private school
fees (see Table 4, Panel B, column 10). Conversely, there is a decline in health expenditures,
which seems to be mostly accounted for by the highest quantiles (quantile plot available on
request). These could be statistical accidents–they go in opposite direction and the effect overall
index is not significant. But the positive impact on school fees is consistent with claims that
microfinance clients are more likely to send their children to private schools (the first sign of an
increase in educational aspirations in India).
5
External validity and multiple inference: Comparing our results with other evaluations of similar programs
External validity and multiple inference are two issues that arise with any sort of evaluation of
a program like microcredit that operates in many very different locations and has the potential
to affect many different outcomes (and thus does not offer a single “litmus test” of success).
The external validity question is straightforward: are the results obtained in a booming city
like Hyderabad any guide to what one would obtain in rural areas, or in another country? One
advantage of microcredit is that while there are several different models, there is a standard “plain
vanilla” model, adopted by Spandana and thousands of other organizations all over the world.
(This features group lending, weekly or monthly repayment, and fixed-term loans usually lasting
close to a year.)27 The second issue is that of multiple inference. Since we are looking at many
different outcomes (as we should, given the nature of the program), there is always the possibility
that the results we obtain are the statistical artifacts of multiple hypothesis testing. For example,
one of our interesting results is the significant decline in temptation goods consumption. But
this is one of several consumption categories and it could just be due to chance (though this
was an hypothesis we formed before the study was started) that it happens to be smaller in the
treatment group.
Fortunately, four RCTs of the same or a very similar model, run by different organizations,
were launched right after ours, and they are all now reporting results at the time of the first
follow-up (after up to two years). These studies have looked at the same outcomes as the ones we
27
Interest rates differ from country to country since they are closely tied to the salaries of the credit officers.
29
look here, and explicitly tested key hypotheses suggested by our results. Thus, when we compare
our results to theirs, we address both issues of external validity and multiple inference, since our
set of results provides a clear (and limited) set of hypotheses that can be tested in their studies
as well.
Here we thus compare our result to four studies.
(1) Crepon et al. (2011)–CDDP–an evaluation of Al Amana’s program in rural Morocco, a
context where there was essentially no access to credit (even informal) whatsoever before the
introduction of the microcredit program, and in which there was no competition between this
MFI and others.
(2) Angelucci et al. (2012)–AKZ–an evaluation of Compartamos’s program in Mexico, implemented in both urban and rural area in the state of Sonora.
(3) Attanasio et al. (2011)–AADFH–an evaluation of XacBank’s MFI program in rural
Mongolia, randomized at the village level. Two versions of the program were tested, a group
liability offer and an individual liability offer. We report the group results here, for comparability.
(4) Augsburg et al. (2012)–AHHM–an evaluation of a EKI in Bosnia, where microcredit and
formal credit were quite developed when the MFI program was offered to treatment clients (who
were randomly selected out of those who had been rejected for a loan). Unlike the previous three
studies, this is an individual-level randomization among people who applied for a loan, and this
is an individual loan product.
Our results and the results of these four studies, for the headline results of our study, are
presented in Table 10. The five papers’ main results are strikingly consistent for most economic
outcomes.
We have already discussed the increase in microfinance access due to the entry of one new
microfinance institution. Access to microfinance from the MFI who participated in the trials is
13 percentage points higher in our study, 10 percentage point higher in CDDP and in AKZ, and
57 percentage points higher in AADFH. The higher first stage in AADFH is not comparable to
what is found in the other studies, since they it used an oversubscription-type design in which
all surveyed households (in both treated and comparison villages) contained a woman who had
explicitly expressed interest in borrowing prior to randomization, knowing that there was a 2/3
30
chance that her village would be randomly assigned to one of the treatment groups.28 Increase in
access to microfinance from any MFI is 8.8 percentage point in our study, 10 percentage points
in CDDP, 24 percentage points in AADFH, and is not reported in AKZ.
For the whole sample, all of the papers find insignificant impact on monthly consumption
and non-durable consumption; the point estimate for non-durable consumption is negative in all
papers, except in AADFH where it is positive but very small and insignificant. All papers find
decline in either temptation goods expenditures, or spending on festivals and parties, or both;
four out of five papers find significant decrease in temptation goods expenditures. The only study
that does not is CDDP, but it does find a significant decline expenditures on festivals and parties
(which we also find–the other studies don’t report this outcome). The main difference between
our study and most others is that we find an increase in the purchase of durable goods used at
home in the last year, which is not present in any of the other studies except AADFH. However,
it is worth nothing that all endlines were conducted more than a year after microfinance started,
while the durable purchase is over the last year: it is thus possible that the asset purchase took
place early on, and is not detected by the survey. Consistent with that, CDDP reports a increase
in the household assets index owned for the median household, as well as an increase in the
number of cows owned, and AADFH finds increases in ownership of VCRs, radios, and large
household appliances.
All of these studies find an expansion in self-employment activities (business, farming or
cattle raising), which shows up either in revenue expansion (positive in all studies, significant in
two of the four studies where it is reported), or in investment in business assets or inventories
when this data has been collected (positive and significant whenever measured, except AADFH).
In all studies, the impact on businesses comes from the intensive margin (investment in existing
businesses, or households investing in more than one business at the same time), rather than
because new households are induced to start an activity.29 In all studies except CDDP, this
expansion does not translate into significantly higher profit on average. It also does not translate
into significant increase of the profits of the median business: in all studies that report it, the
28
There is no comparable result for AHHM, since their design involves randomization among rejected applicants.
The only place where there is an impact on the probability that a household has a business is in AHHM, but
in this case, it is because treatment households are less likely to shut a business (in a context of a crisis), not
because they are more likely to set one up.
29
31
impact is very small and insignificant at the median. However, there is positive and significant in
profit increase in profits the upper tail of businesses (90th or 95th percentile) in all three studies
that report it.
All studies that report it also find an increase in labor supply in the household’s own business,
at least for some members of the households. AKZ do not report labor supply estimates. CDDP
finds a significant increase in in hours worked in own business among prime age adults (20-40).
AHHM finds an increase in labor supply in the household business among teenagers only. Like
our study, these studies also find no increase (and even a significant decrease in the case of
CDDP) in hours of work outside the own business. Unlike our study, there is no increase in
hours worked overall in any these studies.
Interestingly, it is on the social outcomes, where economic theory makes little prediction, that
the four studies have quite different results. While CDDP, like us, finds no impacts on women’s
empowerment and AHHM and AADFH do not even report the effects, AKZ finds an increase in
women’s decision-making power in the treatment group. On education, AKZ and our study find
no impact on enrollment, though by endline 2 we find that households spend more on school fees,
CDDP has a small positive impact on the probability that children are in school, and AHHM
finds a negative impact on the probability that teenagers are in school (and no impact on the
younger children). We find a decline in health expenditures by endline 2, while CDDP finds an
increase in health expenditures.
One explanation for these divergent outcomes is that they may depend more than the basic
economic results on the MFI’s vision and ideology, and less on standard economic forces. For
example, Al Amana in Morocco lends to both men and women; Spandana, while it lends to both,
does not insist that the loan is to finance a woman’s business (or in fact any business). Compartamos’s program, Crédito Mujer, is in principle targeted at women who are micro-entrepreneurs
(although only about half of them are). These outcomes may also be much more influenced by
the context.
32
6
Conclusion
This study–the first and longest running evaluation of the standard group-lending loan product
that has made microfinance known worldwide–yields a number of results that may prompt a
rethinking of the role of microfinance.
The first result is that, in contrast to the claims sometimes made by MFIs and others,
demand for microloans is far from universal. By the end of our three-year study period, only
38% of households borrow from an MFI30 , and this is among households selected based on their
relatively high propensity to take up microcredit. This does not appear to be an anomaly: the
two other randomized interventions that have a similar design (in Morocco and in Mexico) also
find relatively low take-up, while another study in rural South India that focuses specifically on
take-up of microfinance also finds it to be low (Banerjee, Chandrasekhar, Duflo, and Jackson
2012). Perhaps it should not be surprising that most households either do not have a project
with a rate of return of at least 24%, the APR on a Spandana loan, or simply prefer to borrow
from friends, relatives, or moneylenders due to the greater flexibility those sources provide,
despite costs such as higher interest (from moneylenders) or embarrassment (when borrowing
from friends or relatives) (Collins, Morduch, Rutherford, and Ruthven 2009).
For those who choose to borrow, while microcredit “succeeds” in leading some of them to
expanding their businesses (or choose to start a female-owned business), it does not fuel an
escape from poverty based on those small businesses. Monthly consumption, a good indicator of
overall welfare, does not increase for those who had early access to microfinance, neither in the
short run (when we may have foreseen that it would not increase, or perhaps even expected it to
decrease, as borrowers finance the acquisition of household or business durable goods), nor, more
tellingly, in the longer run, after this crop of households have access to microcredit for a while,
and those in the former control group should be the ones tightening their belts. Business profit
does not increase for the vast majority of businesses, although there are significant increases in
the upper tail. This study took place in a dynamic urban environment, in a context of very
high growth. Microcredit seems to have played very little part in it. However, these results
are not specific to this context: similar findings emerge from four other studies run on different
30
The takeup rate is 42% in treatment areas and 33% percent in control areas.
33
continents, in booms and busts, and in both urban and rural contexts.
Furthermore, in the Hyderabadi context, we find that access to microcredit appears to have
no discernible effect on education, health, or women’s empowerment in the short run . In the
longer run (when borrowing rates are the same, but households in the treatment groups have on
average borrowed for longer), there is still no impact on women’s empowerment, and while there
is an apparent increase in money spent on school fees, there is an apparent decline in health
expenditures. The results differ from study to study on these outcomes, but as a whole they
don’t paint a picture of dramatic changes in basic development outcomes for poor families.
Microcredit therefore may not be the “miracle” that it is sometimes claimed to be, although
it does allow some households to invest in their small businesses. One reason may be that the
average business run by this target group is tiny (almost none of them have an employee), not
particularly profitable, and difficult to expand, even in a high-growth context, given the skill
sets of the entrepreneurs and their life situations. And, consistent with theory, the marginal
businesses that get created thanks to microcredit are probably even less profitable and dynamic:
we find that the average new business in a microcredit treatment area less likely to have an
employee than the new business in the control areas, and the median new business is even less
profitable in treatment versus control areas.
Nevertheless, microcredit does affect the structure of household consumption. We see households invest in home durable goods and restrict their consumption of temptation goods and
expenditures on festivals and parties. They continue to do so several years later, and this decrease in not due to a few particularly virtuous households, but seems to be spread across the
sample. Similar declines in these types of expenses are also found in all the other studies. Altered
consumption thus does not seem to be tied to the ideology of a particular MFI.
Microfinance affects labor supply choices as well: households that have access to loans seem
to work harder on their own businesses. Thus, microcredit plays its role as a financial product in
an environment where access not only to credit, but also to saving opportunities, is limited. It
expands households’ abilities to make different intertemporal choices, including business investment. The only mistake that the microcredit enthusiasts may have made is to overestimate the
potential of businesses for the poor, both as a source of revenue and as a means of empowerment
for their female owners.
34
References
Alexander-Tedeschi, G. and D. Karlan (2007). Cross sectional impact analysis: Bias from
dropouts. Yale University working paper.
Angelucci, M., D. Karlan, and J. Zinman (2012, November). Win some lose some? Evidence
from a randomized microcredit program placement experiment by Compartamos Banco.
J-PAL working paper.
Armendáriz, B. and J. Morduch (2005). The Economics of Microfinance. MIT Press: Cambridge, MA.
Armendáriz, B. and J. Morduch (2010). The Economics of Microfinance, Second Edition. MIT
Press: Cambridge, MA.
Attanasio, O., B. Augsburg, R. De Haas, E. Fitzsimons, and H. Harmgart (2011). Group
lending or individual lending? Evidence from a randomised field experiment in Mongolia.
Pub ref: MPRA Paper No. 35439.
Augsburg, B., R. D. Haas, H. Harmgart, and C. Meghir (2012). Microfinance, poverty and
education. IFS working paper.
Banerjee, A. (2013). Microcredit under the microscope: What have we learnt in the last two
decades, what do we need to know? Annual Review of Economics 5.
Banerjee, A., A. Chandrasekhar, E. Duflo, and M. Jackson (2012). The diffusion of microfinance. MIT working paper.
Buera, F. J., J. P. Kaboski, and Y. Shin (2011, June). The macroeconomics of microfinance.
BREAD working paper.
Collins, D., J. Morduch, S. Rutherford, and O. Ruthven (2009). Portfolios of the Poor: How
the World’s Poor Live on Two Dollars a Day. Princeton University Press.
Consultative Group to Assist the Poor (CGAP) (2009). What do we know about the impact
of microfinance? http://www.cgap.org/p/site/c/template.rc/1.26.1306/.
Crépon, B., E. Duflo, F. Devoto, and W. Pariente (2011, March). Impact of microcredit in
rural areas of Morocco: Evidence from a randomized evaluation. J-PAL working paper.
35
de Mel, S., D. McKenzie, and C. Woodruff (2009). Are women more credit constrained? Experimental evidence on gender and microenterprise returns. American Economic Journal:
Applied Economics 1, 1–32.
DiNardo, J., N. M. Fortin, and T. Lemieux (1996). Labor market institutions and the distribution of wages, 1973-1992: A semiparametric approach. Econometrica 64 (5), 1001–44.
Duflo, E. (2003). Grandmothers and granddaughters: Old age pensions and intra-household
allocation in South Africa. World Bank Economic Review 17, 1–25.
Fulford, S. (2011, May). The effects of financial development in the short and long run. Boston
College working paper.
Giné, X. and R. M. Townsend (2004). Evaluation of financial liberalization: A general equilibrium model with constrained occupation choice. Journal of Development Economics 74,
269–307.
Kaboski, J. and R. Townsend (2005). Policies and impact: An analysis of village-level microfinance institutions. Journal of the European Economic Association 3, 1–50.
Kaboski, J. P. and R. M. Townsend (2011, July). A structural evaluation of a large-scale
quasi-experimental microfinance initiative. Econometrica 79, 1357–1406.
Karlan, D. and J. Zinman (2010a). Expanding credit access: Using randomized supply decisions to estimate the impacts. Review of Financial Studies 23, 433–464.
Karlan, D. and J. Zinman (2010b, January). Expanding microenterprise credit access: Using
randomized supply decisions to estimate the impacts in manila. Innovations for Poverty
Action working paper.
Karlan, D. and J. Zinman (2011). Microcredit in theory and practice: Using randomized credit
scoring for impact evaluation. Science 332 (6035), 1278–1284.
Kling, J. R., J. B. Liebman, and L. F. Katz (2007). Experimental analysis of neighborhood
effects. Econometrica 75, 83–119.
Microcredit Summit Campaign (2012). State of the campaign report. Technical report, RESULTS Educational Fund.
MIX Market (2009). Profile for Spandana.
36
Morduch, J. (1998). Does microfinance really help the poor? Evidence from flagship programs
in Bangladesh. Hoover Institution, Stanford U. working paper.
Morduch, J. (1999). The microfinance promise. Journal of Economic Literature 37, 1569–1614.
Pitt, M. and S. Khandker (1998). The impact of group-based credit programs on poor households in Bangladesh: Does the gender of participants matter? Journal of Political Economy 106, 958–996.
Polgreen, L. and V. Bajaj (2010). India microcredit faces collapse from defaults. New York
Times, A5. 11/17/2010.
Roodman, D. and J. Morduch (2010). The impact of microcredit on the poor in Bangladesh:
Revisiting the evidence. NYU Wagner research paper no. 2010-09.
Thomas, D. (1990). Intra-household resource allocation: An inferential approach. Journal of
Human Resources, 635–664.
Townsend, R. M. (1994). Risk and insurance in village India. Econometrica 62, 539–591.
World Bank (2005). World development indicators. http://data.worldbank.org/products/databooks/WDI-2005.
World Bank (2007). World development indicators, Table 4.14: Exchange rates and prices.
37
Figure 1:
Informal borrowing: borrowing from moneylenders, friends and
family, and buying goods on credit.
Confidence intervals are cluster-bootstrapped at the neighborhood
level. For quantiles .05 to .20, confidence intervals are not reported
because the quantile does not vary sufficiently across neighborhoods to bootstrap standard errors. The point estimates are zero
for these quantiles.
Figure 2:
1
Figure 3:
Figure 4:
Old business: business started at least one year before the
survey.
2
Figure 5:
Old business: business started at least one year before the
survey.
Figure 6:
3
Table 1: Baseline Characteristics
PANEL A: Demographics
Household size
Household expenditure (Rs/mo)
Household owns home
Treatment
(1)
5.13
Control
(2)
5.04
Difference
(3)
0.095
(0.092)
[1.78]
[1.67]
5,485
5,208
277
[4,820]
[4,224]
(232)
0.676
0.674
0.002
Obs
(4)
2,440
2,440
2,435
(0.040)
Household rents home
0.288
0.272
0.016
2,435
(0.034)
School attendance (7-11 yrs old)
0.981
0.974
0.007
1,290
(0.010)
School attendance (12-15 yrs old)
0.853
0.856
-0.002
1,135
(0.025)
Working for a wage (Wage Labor /Job Work)
0.410
0.407
0.003
4,460
(0.034)
Business income (business owners only, Rs/mo)
Total household income (Rs/mo)
3,265
3,393
-128
[3,982]
[7,469]
(541)
4,921
4,825
96
[4,818]
[5,861]
(293)
0.684
0.682
650
2,440
PANEL B: Household savings/insurance/shocks
Household with at least 1 outstanding loan
0.002
2,440
(0.029)
Average loan outstanding (Rs)
Average interest rate (monthly)
Loans taken from moneylender
20,228
23,779
-3,551
[39,131]
[145,791]
(3587)
4.017
3.771
0.245
[10.18]
[2.50]
(0.441)
0.498
0.512
-0.014
4,279
3,727
4,249
(0.045)
Loans taken from friends or neighbors
0.252
0.255
-0.003
4,249
(0.039)
Loans taken from family members
0.137
0.129
0.007
4,249
(0.018)
Loans taken from commercial banks
0.028
0.03
-0.002
4,249
(0.007)
Household with a savings account
0.322
0.34
-0.019
2,439
(0.028)
Household with life insurance
0.23
0.237
-0.007
2,440
(0.023)
Household with health insurance
0.003
0.003
0
2,440
(0.002)
Household spent Rs. 500 on health shock, previous year
0.425
0.38
0.045
2,439
(0.033)
Household w/ sick member had to borrow
0.5
0.581
-0.081
774
(0.04)
PANEL C: Business
Number of businesses per household
Households with at least one business
0.301
0.32
-0.019
[0.62]
[0.68]
(0.034)
0.233
0.242
-0.009
2,440
2,440
(0.025)
Business with any employees (%)
0.094
0.056
0.038
735
(0.028)
Business without any assets (%)
0.15
0.157
-0.007
747
(0.044)
Average revenues (Rs/mo)
9,396
10,051
-655
[13,945]
[15,582]
(1242)
695
Note: Standard deviations of nonbinary variables in brackets (cols 1 and 2). Standard errors of differences, clustered at the area level, in
parentheses (col 3). All monetary amounts in 2007 Rs. See Appendix 2 for variable definitions.
Table 2: Descriptive statistics, control households (endline 1 and 2)
EL 1 Control
EL 2 Control
Difference (EL2EL1)
Obs.
(EL 1)
Obs.
(EL 2)
(1)
(2)
(3)
(4)
(5)
2,943
PANEL A: Demographics
Household-level characteristics
Household size (number people)
Household expenditure (Rs/mo)
Household is owner
Household is renter
Waterproof house
Household has a color TV
Household has a cell phone
Total income (Rs/mo)
Business income (business owners only, Rs/mo)
5.640
[ 2.15]
6.269
[2.55]
0.63
(0.037)
3,248
7,662
[5,822]
11,497
[8,732]
3,835
(212)
3,248
2,943
(0.026)
3,263
2,943
0.754
0.644
-0.113
0.089
0.087
0.002
(0.01)
3,264
2,943
0.558
0.679
0.121
(0.020)
3,254
2,941
0.603
0.806
0.203
(0.011)
3,252
2,942
0.642
0.861
0.218
(0.013)
3,253
2,942
4,009
[5,012]
7,735
[6,898]
3,726
(358)
3,248
2,943
2,532
[5,268]
4,991
[6,771]
2,459
(324)
1,105
1,231
Person-level characteristics
School attendance (7-11 yr olds)
School attendance (12-15 yr olds)
Working for a wage (all working individuals)
0.979
0.982
0.003
(0.005)
1,799
1,621
0.881
0.867
-0.014
(0.013)
1,424
1,314
0.589
0.540
-0.049
(0.013)
6,482
6,338
0.887
0.905
0.018
(0.008)
3,264
2,943
0.819
0.848
0.029
(0.053)
1,920
2,943
0.241
0.316
0.074
(0.018)
3,263
2,943
0.122
0.765
0.643
(0.030)
3,263
2,943
0.635
0.530
0.781
0.454
0.146
-0.077
(0.012)
(0.022)
3,264
1,905
2,950
2,121
10,304
PANEL B: Savings/insurance/ shocks
Household-level characteristics
Household with at least 1 outstanding loan
Household with a savings account
Household with life insurance
Household with health insurance
Household spent Rs. 500 on health shock, previous yr
Household w/ sick member had to borrow
Loan-level characteristics
Average loan amount outstanding (Rs)
Average interest rate (mnthly, loans w/ interest rt data)
Loan from moneylender
Loan from friends or neighbors
Loan from family members
Loan from commercial banks
20,914
[57,724]
25,315
[51,348]
4,401
(937)
9,602
2.492
[15.86]
2.437
[8.93]
-0.055
(0.324)
3,362
7,642
0.264
0.231
-0.031
(0.013)
10,407
10,316
0.154
0.111
-0.043
(0.010)
10,407
10,316
0.011
0.003
-0.007
(0.003)
10,407
10,316
0.072
0.026
-0.046
(0.009)
10,407
10,316
2,943
PANEL C: Businesses
Household-level characteristics
Number of businesses per household (all HHs)
Number of female-owned businesses (all HHs)
Households with at least one business (all HHs)
0.436
[0.72]
0.561
[0.79]
0.125
(0.019)
3,234
0.183
[0.49]
0.234
[0.51]
0.052
(0.012)
3,209
2,943
0.077
(0.009)
3,234
2,943
0.342
0.418
0.092
0.104
0.011
(0.009)
1,385
1,636
0.348
0.092
-0.256
(0.032)
1,543
1,652
15,682 [ 42,973]
4,365
(1,891)
1,422
1,619
Business-level characteristics
Business with any employees
Business without any assets
Average revenues (Rs/mo)
11,317
[49,475]
Notes: Summary statistics for comparison areas only. Standard deviations of nonbinary variables in brackets (cols 1 and 2). Standard
errors of differences, clustered at the area level, in parentheses (col 3). All monetary amounts in 2007 Rs. See Appendix 2 for variable
Table 3: Borrowing
Borrows from:
Amount borrowed from:
Spandana
(2)
Informal lender
(3)
A bank
(4)
Any MFI
(5)
Panel A: endine 1
Treatment
0.088***
(0.027)
0.13***
(0.021)
-0.052**
(0.021)
0.0026
(0.012)
Mean in control
Stdev in control
Nobs
0.052
0.22
6811
0.76
0.43
6811
0.067***
(0.020)
0.11
0.31
6142
Any MFI
(1)
0.18
0.39
6811
Panel B: endine 2
Treatment
0.0058
(0.030)
Mean in control
Stdev in control
Obs
Notes:
0.33
0.47
6142
Any MFI
(borrowers only)
Number of cycles
borrowed from an
MFI
Spandana
(7)
Informal lender
(6)
(8)
A bank
(9)
(10)
1355***
(447)
1030
(785)
1391***
(239)
-1072
(2519)
49
(2157)
0.11***
(0.041)
0.079
0.27
6811
2374
6652
6811
12976
10216
1616
597
2907
6811
41045
78033
6811
8422
101953
6811
0.32
0.67
6816
0.0024
(0.018)
0.00042
(0.0085)
869
(690)
2344**
(1052)
1046***
(306)
137
(2922)
-1187
(1081)
0.133*
(0.068)
0.6
0.49
6142
0.073
0.26
6142
5544
11348
6142
16752
14192
2094
1567
5618
6142
32356
76704
6142
6127
40308
6142
0.72
1.09
5926
(1): The table presents the coefficient of a "treatment" dummy in a regression of each variable on treatment (with control variables listed in the text). Cluster-robust standard errors in parentheses. Results are weighted to
account for oversampling of Spandana borrowers.
(2) "Informal lender" includes moneylenders, loans from friends/family, and buying goods/services on credit. Number of loan cycles from an MFI (col 10) is the maximum number of loan cycles borrowed with a single
MFI, including the current loan (if any); number of cycles is zero for MFI never-borrowers.
(3) All monetary amounts in 2007 Rs.
(4) * significant at the 10% level, ** at the 5% level, *** at the 1% level.
Table 4: Consumption
Monthly (per capita)
Total
(1)
Panel A: endine 1
Treatment
10.1
Mean in control
Stdev in control
Nobs
Non
durable
(2)
Temptation goods
(3)
Durable
(total)
(4)
Festivals
(5)
Yearly (total)
Home
Home
repairs
repairs
(mean if
(any>Rs
any>Rs
500)
500)
(6)
(7)
Health
(8)
Education: Education:
total
Fees
(9)
(10)
(37.2)
-6.6
(31.8)
-8.73*
(4.88)
1154*
(682)
-763*
(454)
-0.03
(0.020)
-1613
(3588)
-10
(53)
-6.93
(48.0)
8.37
(31.5)
1419.2
978.3
6827
1304.8
852.4
6781
83.9
130.2
6863
6609
19481
6781
3732
5851
6827
0.51
0.5
6834
18313
65428
2198
630
1916
6827
777
1179
5415
346
679
5404
-44.9
(46.9)
0.0065
1755.2
1209.5
6142
-9.99
(6.64)
0.007
117.7
182.4
6142
62
(524)
205
(205)
0.004
(0.017)
0.0028
0.57
0.5
6141
584
(7039)
-130*
(75)
70
(69)
88**
(42)
PANEL B: endline 2
-48.3
(51.4)
0.0054
1914.3
1354.9
6142
Mean in control
8639
5994
28876
1022
1142
513
Stdev in control
18438
6901
192246
2655
1691
1211
Obs
6140
6103
3439
6141
4910
4910
Notes:
(1): The table presents the coefficient of a "treatment" dummy in a regression of each variable on treatment (with control variables listed in text).
Cluster-robust standard errors in parentheses. Results are weighted to account for oversampling of Spandana borrowers.
(2) See Appendix 2 for description of the construction of the profits, sales, and inputs variables.
(3) All monetary amounts in 2007 Rs.
(4) * significant at the 10% level, ** at the 5% level, *** at the 1% level.
Table 5: Business Creation and outcomes (entire sample)
in the last year
currently
in the last month
Num.
Value of
Num.
female
business
Started a business business Closed a assets
business started
started business acquired
(1)
(2)
(3)
(4)
(5)
Value of Has at
Num.
business least a business
assets business owned
(6)
(7)
(8)
Business Business Business
revenue inputs
profits
(9)
(10)
(11)
Panel A: endine 1
Treatment
0.0093
(0.0061)
0.016**
(0.0075)
0.015***
(0.0054)
0.002
(0.0076)
389*
(212)
606
(383)
0.01
(0.022)
0.022
(0.033)
920
(1181)
244
(1052)
357
(313)
Mean in control 0.047
Stdev in control 0.21
Nobs
6757
0.053
0.25
6757
0.026
0.17
6762
0.037
0.19
2352
280
4038
6800
2498
10802
6800
0.34
0.47
6805
0.44
0.72
6805
4856
33108
6608
4055
30446
6685
745
10695
6239
Panel B: endine 2
Treatment
-0.00049
(0.010)
0.0023
(0.013)
-0.005
(0.0062)
-0.00042
(0.0064)
-134
(208)
1288**
(531)
0.023
(0.023)
0.047
(0.040)
267
(527)
-540
(543)
557
(371)
0.093
0.33
6142
0.047
0.23
6142
0.053
0.23
6142
1007
9623
6142
5003
14423
6142
0.42
0.49
6142
0.56
0.79
6142
5847
16784
6116
5225
20603
6116
953
11280
6090
Mean in control 0.083
Stdev in control 0.28
Obs
6142
Notes:
(1): The table presents the coefficient of a "treatment" dummy in a regression of each variable on treatment (with control variables listed
in the text). Cluster-robust standard errors in parentheses. Results are weighted to account for oversampling of Spandana borrowers.
(2) The outcome variables are set to zero when the household does not have a business.
(3) business outcomes are aggregated at the household level when the households have more than one businesses.
(4) Information on closing a businesses in the year prior to the endline 1 survey was only collected for those who had a business as of
endline 1.
(5) Observations with missing or inconsistent itemized sales or revenues are dropped in columns 9 to 11.
(6) See Appendix 2 for description of the construction of the profits, sales, and inputs variables.
(7) All monetary amounts in 2007 Rs.
(8) * significant at the 10% level, ** at the 5% level, *** at the 1% level.
Table 6: Treatment effect on pre-existing business outcomes
monthly
Profit (Rs.)
(1)
Inputs (Rs.)
(2)
currently
Revenues (Rs.)
(3)
Has any
employees?
(4)
Num. of
employees
(5)
Wages paid
out (Rs per
month)
(6)
Assets used in
business
(7)
Panel A: endine 1
Treatment
2194**
(1105)
1605
(3269)
5356
(3744)
-0.056
(0.084)
0.0058
(0.019)
-117
(156)
855
(1080)
Mean in control 2000
Stdev in control 12315
Obs
1598
12417
51033
1994
14578
47922
1929
0.42
1.74
2054
0.12
0.33
1927
445
3158
2054
6862
17336
2054
Panel B: endine 2
Treatment
1019
(1160)
-3637**
(1709)
-31
(1502)
-0.16
(0.12)
0.0093
(0.019)
-362
(293)
1817
(1759)
15199
36234
1543
15386
24607
1540
0.56
2.93
1559
0.13
0.34
1559
1187
7079
1559
12405
22077
1559
Mean in control 2392
Stdev in control 19878
Obs
1525
Notes:
(1): The table presents the coefficient of a "treatment" dummy in a regression of each variable on treatment (with control variables
listed in the text). Cluster-robust standard errors in parentheses. Results are weighted to account for oversampling of Spandana
borrowers.
(2) The sample is restricted to households who owned a business prior to Spandana's entry.
(3) All monetary amounts in 2007 Rs.
(4) * significant at the 10% level, ** at the 5% level, *** at the 1% level.
Table 7: Treatment + selection effects on new business outcomes (endline 1)
monthly
Profit (Rs.)
(1)
Endline 1
Treatment
-3517
(3802)
Profit (median)
(2)
-1250***
(404)
Inputs (Rs.)
(3)
-4965
(4038)
Revenues (Rs.)
(4)
-8093
(7291)
currently
Has any
employees?
(5)
-0.049*
(0.028)
Num. of
employees
(6)
-0.20*
(0.11)
Assets used in
business
(7)
-812
(2205)
Standardized
outcome: scale
(1 through 7,
except 2)
(8)
-0.096*
(0.055)
Mean in control
6081
1609
12114
17423
0.094
0.29
8411
0.0058
Stdev in control
43517
53020
91782
0.29
1.33
24130
0.7
Obs
270
270
339
332
319
356
356
356
Notes:
(1): The table presents the coefficient of a "treatment" dummy in a regression of each variable on treatment (with control variables listed in the text). Cluster-robust
standard errors in parentheses. Results are weighted to account for oversampling of Spandana borrowers.
(2) The sample is restricted to households who did not own a business prior to Spandana's entry.
(3) The outcome var in col 8 is an average of z-scores of the outcomes in cols 1 and 3-7 (value - control mean)/(control std dev).
(4) All monetary amounts in 2007 Rs.
(5) * significant at the 10% level, ** at the 5% level, *** at the 1% level.
Table 8: Labor supply
Panel A: endline 1
Treatment
Mean in control
Stdev in control
Nobs
Panel B: endline 2
Treatment
Hours worked
Hours worked by head and
by head and
spouse for
spouse, total
wage
(1)
(2)
Hours worked
by head and
Hours worked
spouse, own
by children
business
aged 9-17
(3)
(4)
3.22**
(1.42)
0.44
(1.42)
2.78*
(1.48)
0.19
(0.38)
57.8
35.9
6827
32
34.4
6827
25.8
34.6
6827
3
10.9
3880
1.07
(1.18)
-0.7
(1.48)
1.77
(1.58)
-0.12
(0.30)
Mean in control
51.3
25.9
25.4
2.76
Stdev in control
35.4
31.4
33.4
9.83
Obs
6142
6142
6142
3570
Notes:
(1): The table presents the coefficient of a "treatment" dummy in a regression of each
variable on treatment (with control variables listed in the text). Cluster-robust standard
errors in parentheses. Results are weighted to account for oversampling of Spandana
borrowers.
(2) Column 4 includes only households with children aged 9-17.
(3) * significant at the 10% level, ** at the 5% level, *** at the 1% level.
Table 9: Social outcomes and women's empowerment
Woman
primary
decisionmaker
Woman
primary
decisionmaker
(non-food)
(2)
Woman
primary
decisionmaker on
loans
(3)
Child's
major
illness
0.021
(0.032)
0.014
(0.017)
-0.014
(0.016)
-0.046
(0.034)
-0.016
(0.034)
0.0071
(0.023)
0.66
0.47
6855
0.52
0.50
6855
0.28
0.40
6033
0.28
0.45
3943
1.21
0.84
4062
0.83
0.64
1971
0.00
0.46
6862
0.012
(0.024)
-0.009
(0.023)
0.0037
(0.017)
-0.00033
(0.012)
0.04
(0.033)
0.023
(0.032)
-0.0089
(0.020)
(1)
Panel A: endine 1
Treatment
0.0071
(0.034)
Mean in control
Stdev in control
Nobs
Panel B: endine 2
Treatment
Girls'
education
(4)
Teenage
boys'
education
(5)
Index of
social
outcomes
(6)
(7)
Mean in control
0.61
0.50
0.35
0.39
1.2
0.85
0.00
Stdev in control
0.49
0.50
0.41
0.49
0.82
0.63
0.52
Obs
6142
6142
5562
5942
3592
1776
6142
Notes:
(1): The table presents the coefficient of a "treatment" dummy in a regression of each variable on treatment (with control
variables listed in the text). Cluster-robust standard errors in parentheses. Results are weighted to account for oversampling
of Spandana borrowers.
(2) In column (3) the sample is restricted to households that have taken a loan.
(3) In column (4) the sample is restricted to households with children between the age of 0 and 18.
(4) In column (5) the sample is restricted to households with girls between the age of 4 and 18.
(5) In column (6) the sample is restricted to households with boys between the age of 13 and 18.
(6) The outcome var in col 7 is an average of z-scores of 16 social outcomes; see section 4.4 text for details. Z-score is (value control mean)/(control std dev).
(7) * significant at the 10% level, ** at the 5% level, *** at the 1% level.
Table 10: Findings from RCTs of microfinance
Length of Increase in Consumpfollowup borrowing tion per
from an
capita
MFI
(1)
(2)
(3)
Non
durable
consumption per
capita
(4)
Temptation
goods
and/ or
festivals
(5)
Home use Has a self Investment in Gross
durable
employed self employed revenue
purchase activity
activities
self
employed
activity
(6)
(7)
14 months
Bosnia
(AHHM)
20 percent.
points
negative,
insig
negative,
insig
-12.01%
(p<0.1)
negative,
insig
Mexico
(AKZ)
increase of
12.74%
(p<0.01) in
total no.
loans
received in
last 2 yrs;
increase of
18.61%
(p<0.01) in
total loan
amt
received in
last 2 yrs
positive,
insig effect
on "nights
did not go
hungry"
negative,
insig for
"Amount
spent
on
groceries"
decrease of 5.91%,
(p<0.05) on
amount
spent on
temptation
goods;
negative
insig for
spending
on family
events
negative
negative
insig (Made insig
home
improvement)
increase in
credit
access:
495.24%
(p<0.05),
increase in
loan
amount
479.77%
(p<0.05)
negative,
insig for
total
consumption
negative
insig
negative,
negative,
insig for sin; insig
-5.55%
(p<0.10) for
"social"
2-3 years
Morocco 2 years
(CDDP)
(cont.)
9.31%
(p<0.1)
no effect on
business as
main
income
source
(8)
26.5% (p<0.01)
(9)
Profit of
self
employed
activity:
Median
(10)
Profit of
Women's
self
empower
employed ment
activity:
90th
percentile
(11)
(12)
(13)
Education Labor
supply
overall
(14)
Labor
supply,
own
business
(15)
(16)
positive,
insig
N.A
N.A
N.A
-9.57% for
school
attendance
ages: 16-19
positive,
insig
542.22%
(p<0.05) for
16-19 year
olds, insig
overall
negative,
insig
small and
insig
positive, sig
(p<10),
value not
specified
increase of
0.82%
(p<0.01) in
Participates
in any
financial
decisions
increase of
in # of
household
issues has a
say on
positive
insig for
"Amount
spent on
school and
medical
expenses"
negative
insig for
participate
in an
economic
activity
N.A
positive,
positive
insig
insig for
"global
monetary
revenues";
35.87%
(p<0.01) for
agriculture
and 11.10%
(p<0.1) for
livestock
small and
insig
increase of insig
2.3: 57.57%
(p<0.15)
neg insig
for
schooling
expend.
insig
increase of
8.1% (p<.10)
for
agriculture,
positive
insig for
livestock
activities,
negative
insig for
business
positive,
insig
increase of
increase of
36.25% (p<0.05) 26.73%
(p<0.05)
for variable
inputs; not
reported for
fixed
investment
positive, insig
for fixed
investment,
13.18% (p<0.05)
for variable
inputs
(expenses)
Profit of
self
employed
activity :
Mean
Table 10: Findings from RCTs of microfinance (con't)
Length of Increase in Consump- Non
followup borrowing tion per
durable
from an
capita
consumpMFI
tion per
capita
(1)
(3)
(4)
increase of positive
48% in the insig
probability
of receiving
microcredit
positive
insig
negative
negative
significant insig
for
cigarettes
consumptio
n (p<0.10)
India
(BDGK)
Endline 1:
increase of
48.3% from
any MFI
(p<0.01)
and
257.25%
from
Spadana
(p<0.01);
endline 2:
positive
insig
negative,
insig
(endline 1
and 2)
Endline 1: 10.41% for
temptation
goods
(p<0.10);
20.46% for
festivals
(p<0.10);
endline 2
insig
positive
insig
(endline 1
and 2)
(5)
Home use Has a self Investment in Gross
durable
employed self employed revenue
purchase activity
activities
self
employed
activity
Mongolia 1.5 years
1.5 years
MFI
exposure
(endline 1)
and 3 years
vs. 1.5 years
MFI
exposure
(endline 2)
(2)
Temptation
goods
and/ or
festivals
(6)
Endline 1:
17.46% for
durables
(total)
(p<0.10);
negative
insig for
home
repairs;
endline 2
insig
(7)
(8)
(9)
positive
insig for
probability
of any type
of business
positive, insig N.A
for probability
of tools,
probability of
unsold stock
and raw
materials and
probability of
riding
equipment.
Negative, insig
for number of
cattle and
number of
animals.
Endline 1:
positive
insig for
stated any
business;
3% for
number of
business
started
(p<0.05)
Endline 1:
138.88% for
value of
business assets
acquired
(p<0.10)
positive,
insig
(endline 1
and 2)
Profit of
self
employed
activity :
Mean
Profit of
self
employed
activity:
Median
(10)
Profit of
Women's
self
empower
employed ment
activity:
90th
percentile
(11)
(12)
Education Labor
supply
overall
(13)
(14)
increase of N.A
16.20%
(p<0.10) in
the
probability
of female
business
negative,
insig
N.A
N.A
positive,
insig
(endline 1
and 2)
insig
(endline 1
and 2)
positive
postivie,
insig
insig,
endline 1;
increase of
20% (p<.05)
endline 2
(for all
businesses)
17.21% for
education
fees in
endline 2
(p<0.05),
insig in
endline 1
Labor
supply,
own
business
(15)
(16)
insig
positive,
insig
Endline 1:
increase of
5.58%
(p<0.05);
endline 2
insig
Endline 1:
increase of
11% (p<.05),
endline 2
insig
Notes: (1) Bosnia results are from Augsburg, De Haas, Harmgart and Meghir (2012); Mexico results are from Angelucci, Karlan and Zinman (2012); Morocco results are form Crépon, Devoto, Duflo and
Parienté (2011). India results are from Banerjee et al. (2012).
(2): Effect sizes for India, Mexico and Morocco studies are percentages of endline control group means. Effect sizes for Bosnia study are percentages of overall baseline. Effect sizes for Mongolia
study are percentages of control baseline means (when available).
Table A1: Treatment-Control balance in fixed characteristics (Endline 1)
Spouse
Prime-
Treatment
Spouse is
literate
(1)
-0.011
(0.021)
Control Mean
0.54
Control Std Dev 0.5
Obs
6139
works
for a
wage
(2)
-0.011
(0.024)
aged
Household women
size
(18-45)
(3)
(4)
-0.032
-0.022
(0.083)
(0.026)
Any teen
(13-18)
in HH
(5)
0.023
(0.016)
Old
businesses
owned
(6)
0.006
(0.030)
Adult
Own land, Own land, labor
Hyderabad village
supply
(7)
(8)
(9)
-0.0028
0.0045
-0.28
(0.0069)
(0.027)
(2.17)
Adult
labor
supply, HH
business
(10)
0.74
(1.31)
0.23
0.42
6229
5.64
2.15
6827
0.49
0.5
6862
0.38
0.67
6762
0.061
0.24
6830
18.8
35.5
6827
1.46
0.82
6862
0.19
0.4
6819
88.1
58.5
6827
Note: The table presents the coefficient of a "treatment" dummy in a regression of each variable on treatment (with no control variables).
Cluster-robust standard errors in brackets. Results are weighted to account for oversampling of Spandana borrowers. Spouse is the wife
of the household head, if the head is male, or the household head if female. Household size is the total number of household members
(not adult equivalents). An old business is a business started at least 1 year before the survey. Adult labor supply is total work hours by
HH members aged 18-60. * significant at the 10% level, ** at the 5% level, *** at the 1% level.
Table A2: Endline 2 Attrition
Panel A: Attrition in treatment vs. control
Found in endline 2, in treatment
Found in endline 2, in control
p-value of difference
0.8889
0.9017
0.248
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Exp per
capita
Temptation goods
Durables
Festival
spending
Has MFI
loans
Old
business
New
business
Business
profit
Panel B: Attrition and household characteristics (endline 1)
Attrited
130.9***
(42.4)
0.032
(21.2)
1702.8
(1273.9)
188.9
(197.6)
-0.055***
(0.015)
-0.015
(0.031)
-0.015*
(0.0083)
751.3
(882.3)
Non-attriter mean
Obs
1423.2
6827
348
6827
7075.8
6781
3477.3
6827
0.24
6811
0.39
6762
0.063
6757
890.6
6239
Panel C: Attrition and household characteristics (endline 1) in treatment vs. control
Attrited X treatment
-18.8
40
-110.9
-91.7
-0.090***
(82.7)
(41.8)
(2429.5)
(388.5)
(0.028)
-0.069
(0.061)
0.0016
(0.016)
820.6
(1584.8)
Non-attriter mean, C
Obs
0.38
6762
0.054
6757
718.4
6239
1405.4
6827
369.9
6827
6439.3
6781
3707.4
6827
0.18
6811
Notes:
(1): Panel B presents the coefficient from regressing a dummy for "attrited between endline 1 and endline 2" on various endline
1 characteristics, to compare attritors and non attritors.
(2): Panel C investigates whether the characteristics of th attritors is different in treatment and control. The regression controls
for the main effects of attrition and of treatment (coefficients not reported).
(3) All monetary amounts in 2007 Rs.
(4) * significant at the 10% level, ** at the 5% level, *** at the 1% level.
Table A3: Industries of old and new businesses
Old
business,
treatment
(1)
Old
business,
control
(2)
Food/agriculture
0.227
0.243
Clothing/sewing
0.210
0.186
Rickshaw/driving
0.103
0.103
Repair/construction 0.042
0.052
Crafts/vendor
0.020
0.029
Other
0.397
0.380
Nobs
Notes:
1424
1261
Treatmentcontrol
difference
(3)
-0.017
[0.028]
0.024
[0.020]
0.000
[0.021]
-0.010
[0.010]
-0.010
[0.008]
0.018
[0.042]
New
business,
treatment
(4)
New
business,
control
(5)
0.299
0.214
0.135
0.185
0.056
0.110
0.016
0.035
0.024
0.040
0.470
0.416
251
173
Treatmentcontrol
difference
(6)
0.085*
[0.044]
-0.050
[0.033]
-0.054*
[0.028]
-0.019
[0.015]
-0.017
[0.017]
0.054
[0.056]
Old (new) businesses are those started more (less) than 1 year before the survey. Cluster-robust standard
errors in brackets. Results are weighted to account for oversampling of Spandana borrowers. * significant at
the 10% level, ** at the 5% level, *** at the 1% level.
Appendix 1: Theoretical appendix
Consider the decision to buy a durable. First take a case where the consumer does not buy
the durable in period 1. The condition for this is simple. Assuming that if she is indifferent she
buys the durable the condition is:
max0≤b≤bmax {u(y + b) + δu(y − rb)} > max0≤b≤bmax {u(y − (1 − a)cd + b) + δu(y + acd − rb)}
Now consider the decision in period s > 1,assuming that in all previous periods the durable was
not purchased. The only possible difference with period 1 is that they might have borrowed in
the past. If they have not it is the same problem and has the same answer. The interesting case
is when they have borrowed an amount b−1 > 0in the previous period. The question therefore
comes down to whether
u(y − rb−1 ) + δu(y) > u(y − (1 − a)cd − rb−1 ) + δu(y + acd ).
By the concavity of u
u(y − rb−1 ) − u(y − (1 − a)cd − rb−1 ) > u(y + b0 ) − u(y − (1 − a)cd + b0 )
for any b0 ≥ 0.
Now let b0 = argmax0≤b≤bmax {u(y + b) + δu(y − rb)} ≥ 0. Then from above
u(y + b0 ) + δu(y − rb0 ) > u(y − (1 − a)cd + b0 ) + δu(y + acd − rb0 )
which can be rewritten to say
u(y + b0 ) − u(y − (1 − a)cd + b0 ) > δu(y + acd − rb0 ) − δu(y − rb0 ).
1
Finally by the concavity of u,
δu(y + acd − rb0 ) − δu(y − rb0 ) ≥ δu(y + acd ) − δu(y).
Combining these inequalities we end up with
u(y − rb−1 ) − u(y − (1 − a)cd − rb−1 ) > δu(y + acd ) − δu(y),
which is exactly what we needed to show that there is no durable purchase in period s.
In other words, if there are no durable purchases in period 1, there are none in any subsequent
period.
Conversely, if there is a period s such that the borrower has not purchased the durable or
borrowed in period s − 1 and does not plan to borrow in period s + 1,then that period just like
period 1, and she should make exactly the same choice as in period 1 in period s. So if she buys
the durable in period 1, she should buy it in period s. If she borrows b in period 1, she should
borrow the same amount in period s.
Among other things, this tells us that if there is no borrowing in period 2, then the decision
in period 1 will be reproduced in every odd period until the first period 2m + 1 (such that m > 0)
where the person borrows in period 2m + 2.
To complete the argument, we need to rule out this last possibility. Consider the first pair
of periods, (2m + 1, 2m + 2) where the consumer borrows in the even period. This means that
2m + 2 is not the last period but also that she does not borrow in period 2m + 3. Here there
are four possible scenarios; in one she buys the durable in both period 2m + 1 and 2m + 3, one
in which she borrows in neither and one each where she buys the durable in one of those two
periods. In the first case, her utility from periods 2m + 1 to 2m + 3 will be
u(y − (1 − a)cd ) + δu(y + acd + b) + δ 2 u(y − (1 − a)cd − rb),
where b > 0 is the amount she borrows in period 2m + 2. Consider the alternative plan where
she borrows the same amount, but in period 2m + 1, and does not borrow in period 2m + 3.
2
Then her utility from the same three periods will be
u(y − (1 − a)cd + b) + δu(y + acd − rb) + δ 2 u(y − (1 − a)cd ).
Now first consider the decision to borrow in period 2m + 2. A necessary condition for this
borrowing is that
u0 (y + acd ) > δru0 (y − (1 − a)cd ),
which implies that δr has to be less than 1.
Next consider the expression
u(y−(1−a)cd +b)+δu(y+acd −rb)+δ 2 u(y−(1−a)cd )−u(y−(1−a)cd )−δu(y+acd +b)−δ 2 u(y−(1−a)cd −rb),
which using the Intermediate Value Theorem can be rewritten as
bu0 (y1 ) + δ 2 rbu0 (y3 ) − δ(b + rb)u0 (y2 )
where
y1 ∈ (y − (1 − a)cd , y + b − (1 − a)cd )
y2 ∈ (y + acd , y + acd + b)
y3 ∈ (y − (1 − a)cd − rb, y − (1 − a)cd )
It is clear that y3 ≤ y1 and y2 ≥ y1 (the latter because b < cd ) and therefore
bu0 (y1 ) + δ 2 rbu0 (y3 ) − δ(b + rb)u0 (y2 ) ≥ bu0 (y1 )[1 − δ − δr + δ 2 r] > 0
as long as δr < 1. Hence the original plan cannot have been a maximum. Analogous arguments
can be made in all the other cases.
This concludes the argument. Every pair of periods will be like the first pair. We just need
3
to determine the choice in the first two periods assuming that there are no further periods and
then apply to all future pairs of periods.
Appendix 2: Variable definitions
Go to http://www.povertyactionlab.org/projects/project.php?pid=44 to download the survey
instruments
(both
in
English
and
in
Telugu).
Business variables
Business: The survey defined a business as follows: “each business consists of an activity you
conduct to earn money, where you are not someone’s employee. Include only those household
businesses for which you are either the sole owner or for which you have the main responsibility.
Include outside business for which you are the person in the household with the most responsibility.” Households who indicated that they owned a business were asked to answer a questionnaire
about each business. The person in the household with the most responsibility for the business
answered the questions about that business.
All variables reported in the paper are at the household level, i.e. if a household owns multiple
businesses, the values for each business are summed to calculate a household-level total.
Business revenues: Respondents were asked: “For each item you sold last month, how
much of the item did you sell in the last month, and how much did you get for them?” The
respondent was asked to list inputs one by one. They were also asked for an estimate of the total
revenues for the business. If the itemized total and the overall total did not agree, respondents
were asked to go over the revenues again and make and changes, and/or change the estimate of
the total revenues for the business last month.
Business inputs: Respondents were asked: “How much did you pay for inputs (excluding
electricity, water, taxes) in the last day/week/month, e.g. clothes, hair, dosa batter, trash,
petrol/diesel etc.? Include both what was bought this month and what may have been bought
at another time but was used this month. List all inputs and then list total amount paid for
each input. Do not include what was purchased but not used (and is therefore stock), i.e. if
you purchased five saris this months but sold only four, then we need to record the purchase
4
price of four saris, not five.” The respondent could give a daily, weekly, or monthly number. All
responses were then converted to monthly.
The respondent was asked to list inputs one by one. They were also asked for an estimate of
the total cost of inputs for the business. If the itemized total and the overall total did not agree,
they were asked to go over the inputs again and make and changes, and/or change the estimate
of the total cost of inputs for the business last day/week/month.
Respondents were asked about electricity, water, rent and informal payments. If they had
not included them previously, these costs were added.
Business profits: Computed as monthly business revenues less monthly business input
costs.
Employees: Respondents were asked: “How many employees does the business have? (Employees are individuals who earn a wage for working for you. Do not include household members).”
Expenditure
Expenditure comes from the household survey, which was answered by the person “who (among
the women in the 18-55 age group) knows the most about the household finances.” Respondents
were asked about “expenditures that you had last month for your household (do not include
business expenditures)” in categories of food (cereals, pulses, oil, spices, etc.), fuel, and 16
categories of miscellaneous goods and services. They were asked annual expenditure for school
books and other educational articles (including uniforms); hospital and nursing home expenses;
clothing (including festival clothes, winter clothes, etc.) and gifts; and footwear.
Per capita expenditure is total expenditure per adult equivalent. Following the conversion
to adult equivalents used by Townsend (1994) for rural Andhra Pradesh and Maharastra, the
weights are: for adult males, 1.0; for adult females, 0.9. For males and females aged 13-18, 0.94,
and 0.83, respectively; for children aged 7-12, 0.67 regardless of gender; for children 4-6, 0.52;
for toddlers 1-3, 0.32; and for infants 0.05. Using a weighting that accounts for within-household
economies of scale does not affect the results (results available on request).
Expenditure: Sum of monthly spending on all goods where monthly spending was recorded,
and 1/12 of the sum of annual spending on all goods where annual spending was recorded.
5
Non-durable expenditure: Total expenditure minus spending on assets (see below).
“Temptation goods”: Sum of monthly spending on meals or snacks consumed outside the
home; pan, tobacco and intoxicants; and lottery tickets/gambling.
Assets
Assets information comes from the household survey, which was answered by the person “who
(among the women in the 18-55 age group) knows the most about the household finances.”
Respondents were asked about 41 types of assets (TV, cell phone, clock/watch, bicycle, etc.): if
the household owned any, how many; if any had been sold in the past year (for how much); if
any had been bought in the past year (for how much); and if the asset was used in a household
business (even if it was also used for household use).
Assets expenditure (monthly): Total of all spending in the past year on assets, divided
by 12.
Business assets expenditure (monthly): Total of all spending in the past year on assets
which are used in a business (even if also used for household use), divided by 12.
6
Descargar